AI Generator Website

AI Generator Website — independent reviews, comparisons, pricing and step-by-step guides on Aizhi.

• 

  Jais (language model)

  Jais is an open-source large language model launched in August 2023. Developed as a collaboration between Emirati AI company G42, the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), and US-based Cerebras Systems, Jais was designed to produce high-quality Arabic text and was also trained on English data. The model's creation was motivated by the underrepresentation of the Arabic language in the field of generative artificial intelligence. It aims to provide a more culturally and linguistically accurate model for the world's 400 million Arabic speakers. Its name is a reference to Jebel Jais, the highest mountain in the UAE. == Background and development == Jais was developed in response to the limited availability of advanced generative artificial intelligence models for the Arabic language, despite it being spoken by over 400 million people. Existing models were often trained on limited or low-quality Arabic web content, resulting in poor performance. The project represents a significant investment by the United Arab Emirates in the field of AI as part of its national strategy. The model was created through a partnership between Inception (now Core42), a subsidiary of the Abu Dhabi-based AI company G42; the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI); and Cerebras Systems, a US company specializing in AI hardware. The model is named after Jebel Jais, the highest peak in the UAE. == Training == The initial version of Jais released in August 2023 had 13 billion parameters. In November 2023, Core42 released Jais 30B, an improved version with 30 billion parameters. Both models were trained on a subset of the Cerebras Condor Galaxy 1 supercomputer. The training dataset consisted of a mix of Arabic, English, and computer code. According to Timothy Baldwin, a professor of natural language processing at MBZUAI, training the model on a diverse Arabic dataset allows it to switch between dialects. == Features == Jais is designed to generate text in both English and Arabic. The project has also released instruction-tuned "Chat" variants for both the 13B and 30B models, which are specifically optimized for conversational applications. Additional functionality for working with images, graphs, and tabular data is planned for future releases.

  Read more →
• 

  Convolutional neural network

  A convolutional neural network (CNN) is a type of feedforward neural network that learns features via filter (or kernel) optimization. This type of deep learning network has been applied to process and make predictions from many different types of data including text, images and audio. CNNs are the de-facto standard in deep learning-based approaches to computer vision and image processing, and have only recently been replaced—in some cases—by newer architectures such as the transformer. Vanishing gradients and exploding gradients, seen during backpropagation in earlier neural networks, are prevented by the regularization that comes from using shared weights over fewer connections. For example, for each neuron in the fully-connected layer, 10,000 weights would be required for processing an image sized 100 × 100 pixels. However, applying cascaded convolution (or cross-correlation) kernels, only 25 weights for each convolutional layer are required to process 5x5-sized tiles. Higher-layer features are extracted from wider context windows, compared to lower-layer features. Some applications of CNNs include: image and video recognition, recommender systems, image classification, image segmentation, medical image analysis, natural language processing, brain–computer interfaces, and financial time series. CNNs are also known as shift invariant or space invariant artificial neural networks, based on the shared-weight architecture of the convolution kernels or filters that slide along input features and provide translation-equivariant responses known as feature maps. Counter-intuitively, most convolutional neural networks are not invariant to translation, due to the downsampling operation they apply to the input. Feedforward neural networks are usually fully connected networks, that is, each neuron in one layer is connected to all neurons in the next layer. The "full connectivity" of these networks makes them prone to overfitting data. Typical ways of regularization, or preventing overfitting, include: penalizing parameters during training (such as weight decay) or trimming connectivity (skipped connections, dropout, etc.) Robust datasets also increase the probability that CNNs will learn the generalized principles that characterize a given dataset rather than the biases of a poorly-populated set. Convolutional networks were inspired by biological processes in that the connectivity pattern between neurons resembles the organization of the animal visual cortex. Individual cortical neurons respond to stimuli only in a restricted region of the visual field known as the receptive field. The receptive fields of different neurons partially overlap such that they cover the entire visual field. CNNs use relatively little pre-processing compared to other image classification algorithms. This means that the network learns to optimize the filters (or kernels) through automated learning, whereas in traditional algorithms these filters are hand-engineered. This simplifies and automates the process, enhancing efficiency and scalability overcoming human-intervention bottlenecks. == Architecture == A convolutional neural network consists of an input layer, hidden layers and an output layer. In a convolutional neural network, the hidden layers include one or more layers that perform convolutions. Typically this includes a layer that performs a dot product of the convolution kernel with the layer's input matrix. This product is usually the Frobenius inner product, and its activation function is commonly ReLU. As the convolution kernel slides along the input matrix for the layer, the convolution operation generates a feature map, which in turn contributes to the input of the next layer. This is followed by other layers such as pooling layers, fully connected layers, and normalization layers. Here it should be noted how close a convolutional neural network is to a matched filter. === Convolutional layers === In a CNN, the input is a tensor with shape: (number of inputs) × (input height) × (input width) × (input channels) After passing through a convolutional layer, the image becomes abstracted to a feature map, also called an activation map, with shape: (number of inputs) × (feature map height) × (feature map width) × (feature map channels). Convolutional layers convolve the input and pass its result to the next layer. This is similar to the response of a neuron in the visual cortex to a specific stimulus. Each convolutional neuron processes data only for its receptive field. Although fully connected feedforward neural networks can be used to learn features and classify data, this architecture is generally impractical for larger inputs (e.g., high-resolution images), which would require massive numbers of neurons because each pixel is a relevant input feature. A fully connected layer for an image of size 100 × 100 has 10,000 weights for each neuron in the second layer. Convolution reduces the number of free parameters, allowing the network to be deeper. For example, using a 5 × 5 tiling region, each with the same shared weights, requires only 25 neurons. Using shared weights means there are many fewer parameters, which helps avoid the vanishing gradients and exploding gradients problems seen during backpropagation in earlier neural networks. To speed processing, standard convolutional layers can be replaced by depthwise separable convolutional layers, which are based on a depthwise convolution followed by a pointwise convolution. The depthwise convolution is a spatial convolution applied independently over each channel of the input tensor, while the pointwise convolution is a standard convolution restricted to the use of 1 × 1 {\displaystyle 1\times 1} kernels. === Pooling layers === Convolutional networks may include local and/or global pooling layers along with traditional convolutional layers. Pooling layers reduce the dimensions of data by combining the outputs of neuron clusters at one layer into a single neuron in the next layer. Local pooling combines small clusters, tiling sizes such as 2 × 2 are commonly used. Global pooling acts on all the neurons of the feature map. There are two common types of pooling in popular use: max and average. Max pooling uses the maximum value of each local cluster of neurons in the feature map, while average pooling takes the average value. === Fully connected layers === Fully connected layers connect every neuron in one layer to every neuron in another layer. It is the same as a traditional multilayer perceptron neural network (MLP). Each neuron in the fully connected layer receives input from all the neurons in the previous layer. These inputs are weighted and summed with the corresponding biases, and then passed through an activation function to perform a nonlinear transformation, generating the output. The flattened matrix goes through a fully connected layer to classify the images. === Receptive field === In neural networks, each neuron receives input from some number of locations in the previous layer. In a convolutional layer, each neuron receives input from only a restricted area of the previous layer called the neuron's receptive field. Typically the area is a square (e.g. 5 by 5 neurons). Whereas, in a fully connected layer, the receptive field is the entire previous layer. Thus, in each convolutional layer, each neuron takes input from a larger area in the input than previous layers. This is due to applying the convolution over and over, which takes the value of a pixel into account, as well as its surrounding pixels. When using dilated layers, the number of pixels in the receptive field remains constant, but the field is more sparsely populated as its dimensions grow when combining the effect of several layers. To manipulate the receptive field size as desired, there are some alternatives to the standard convolutional layer. For example, atrous or dilated convolution expands the receptive field size without increasing the number of parameters by interleaving visible and blind regions. Moreover, a single dilated convolutional layer can comprise filters with multiple dilation ratios, thus having a variable receptive field size. === Weights === Each neuron in a neural network computes an output value by applying a specific function to the input values received from the receptive field in the previous layer. The function that is applied to the input values is determined by a vector of weights and a bias (typically real numbers). Learning consists of iteratively adjusting these biases and weights. The vectors of weights and biases are called filters and represent particular features of the input (e.g., a particular shape). A distinguishing feature of CNNs is that many neurons can share the same filter. This reduces the memory footprint because a single bias and a single vector of weights are used across all receptive fields that share that filter, as opposed to each receptive field having its own bias and vector

  Read more →
• 

  Realization (linguistics)

  In linguistics, realization is the process by which some kind of surface representation is derived from its underlying representation; that is, the way in which some abstract object of linguistic analysis comes to be produced in actual language. Phonemes are often said to be realized by speech sounds. The different sounds that can realize a particular phoneme are called its allophones. Realization is also a subtask of natural language generation, which involves creating an actual text in a human language (English, French, etc.) from a syntactic representation. There are a number of software packages available for realization, most of which have been developed by academic research groups in NLG. The remainder of this article concerns realization of this kind. == Example == For example, the following Java code causes the simplenlg system [2] to print out the text The women do not smoke.: In this example, the computer program has specified the linguistic constituents of the sentence (verb, subject), and also linguistic features (plural subject, negated), and from this information the realiser has constructed the actual sentence. == Processing == Realisation involves three kinds of processing: Syntactic realisation: Using grammatical knowledge to choose inflections, add function words and also to decide the order of components. For example, in English the subject usually precedes the verb, and the negated form of smoke is do not smoke. Morphological realisation: Computing inflected forms, for example the plural form of woman is women (not womans). Orthographic realisation: Dealing with casing, punctuation, and formatting. For example, capitalising The because it is the first word of the sentence. The above examples are very basic, most realisers are capable of considerably more complex processing. == Systems == A number of realisers have been developed over the past 20 years. These systems differ in terms of complexity and sophistication of their processing, robustness in dealing with unusual cases, and whether they are accessed programmatically via an API or whether they take a textual representation of a syntactic structure as their input. There are also major differences in pragmatic factors such as documentation, support, licensing terms, speed and memory usage, etc. It is not possible to describe all realisers here, but a few of the emerging areas are: Simplenlg [3]: a document realizing engine with an api which intended to be simple to learn and use, focused on limiting scope to only finding the surface area of a document. KPML [4]: this is the oldest realiser, which has been under development under different guises since the 1980s. It comes with grammars for ten different languages. FUF/SURGE [5]: a realiser which was widely used in the 1990s, and is still used in some projects today OpenCCG [6]: an open-source realiser which has a number of nice features, such as the ability to use statistical language models to make realisation decisions.

  Read more →
• 

  Fred (chatbot)

  Fred, or FRED, was an early chatbot written by Robby Garner. == History == The name Fred was initially suggested by Karen Lindsey, and then Robby jokingly came up with an acronym, "Functional Response Emulation Device." Fred has also been implemented as a Java application by Paco Nathan called JFRED Archived 2008-08-24 at the Wayback Machine. Fred Chatterbot is designed to explore Natural Language communications between people and computer programs. In particular, this is a study of conversation between people and ways that a computer program can learn from other people's conversations to make its own conversations. Fred used a minimalistic "stimulus-response" approach. It worked by storing a database of statements and their responses, and made its own reply by looking up the input statements made by a user and then rendering the corresponding response from the database. This approach simplified the complexity of the rule base, but required expert coding and editing for modifications. Fred was a predecessor to Albert One, which Garner used in 1998 and 1999 to win the Loebner Prize.

  Read more →
• 

  Qapital

  Qapital is a personal finance mobile application (app) for the iOS and Android operating systems, developed by Qapital, LLC. The app is designed to motivate users to save money through a gamification of their spending behavior. It moves money from a user's checking account to a separate Qapital account, when certain rules are triggered. Its database is used by psychology professor Dan Ariely to study consumer behavior. Qapital was released in Sweden in 2013, then in the US in early 2015. The application was later withdrawn from the Swedish market in April 2015, in order to focus on the US market. == History == The idea for Qapital was conceived by ex-bankers in Sweden. The software was designed by twin brothers Daniel and Andreas Källbom of Studio Källbom and released in Sweden in December 2013. The original software was a personal finance dashboard, similar to Mint.com, to show its users how they spent their money. Qapital introduced the app into the US market with a different design in 2014 and started focusing exclusively on the US market. The app was re-designed to focus on building savings rather than managing personal finances. The Swedish version shut down in April 2015. The app was initially restricted to the iOS platform, but an Android version was released at the end of 2015. Shortly after its US launch, Qapital invited psychology professor Dan Ariely to join its team as its "chief behavioral economist". He uses the app's database to conduct research into behavioral economics and Qapital in turn uses Ariely's research in design and programming decisions. In 2017, Qapital added checking and debit card services to the app. == Concept and features == Qapital is a free personal finance app for iOS and Android devices, intended to encourage its users to save money. Qapital directs each of its users to set savings goals, then automatically transfers money from their checking account to an account for savings, when a rule established in the app is met. It uses the "if this then that" (IFTTT) rule-based web-service. For example, one rule could be that if a user purchases a cup of coffee, then the app will round up the charge to the nearest dollar and deposit the difference into savings. Users connect their bank accounts to Qapital, so it knows when purchases are made. When a rule is met, money for savings are transferred to a Qapital account operated in partnership with Lincoln Savings Bank. As of 2015, Qapital can connect to more than 180 other apps, such as Facebook, Twitter, Dropbox and Instagram. For example, connecting to Jawbone allows the user to set a rule that if they take a certain number of steps during the day, a set amount of money is transferred to savings. The app also allows users to monitor activity among their other financial accounts, such as deposits and withdrawals. == Reception == In an October 2015 review, PC Magazine gave Qapital four out of five marks and an editor rating of "excellent." The review praised the app for having a "lovely design" and criticized it for being a, "bit simplistic in some of its rules." Bankrate, in a May 2015 review, gave the app a score of 3/5 for "ease of use," 5/5 for "features," 4/5 for "effectiveness," 4/5 for "value," for a total score of 16/20. The reviewer criticized Qapital's savings account for providing a low-interest rate, but concluded that its numerous features make the app "intriguing" and "it would be difficult to find a standard bank app more fun to use than Qapital."

  Read more →
• 

  Automatic acquisition of sense-tagged corpora

  The knowledge acquisition bottleneck is perhaps the major impediment to solving the word-sense disambiguation (WSD) problem. Unsupervised learning methods rely on knowledge about word senses, which is barely formulated in dictionaries and lexical databases. Supervised learning methods depend heavily on the existence of manually annotated examples for every word sense, a requisite that can so far be met only for a handful of words for testing purposes, as it is done in the Senseval exercises. == Existing methods == Therefore, one of the most promising trends in WSD research is using the largest corpus ever accessible, the World Wide Web, to acquire lexical information automatically. WSD has been traditionally understood as an intermediate language engineering technology which could improve applications such as information retrieval (IR). In this case, however, the reverse is also true: Web search engines implement simple and robust IR techniques that can be successfully used when mining the Web for information to be employed in WSD. The most direct way of using the Web (and other corpora) to enhance WSD performance is the automatic acquisition of sense-tagged corpora, the fundamental resource to feed supervised WSD algorithms. Although this is far from being commonplace in the WSD literature, a number of different and effective strategies to achieve this goal have already been proposed. Some of these strategies are: acquisition by direct Web searching (searches for monosemous synonyms, hypernyms, hyponyms, parsed gloss' words, etc.), Yarowsky algorithm (bootstrapping), acquisition via Web directories, and acquisition via cross-language meaning evidences. == Summary == === Optimistic results === The automatic extraction of examples to train supervised learning algorithms reviewed has been, by far, the best explored approach to mine the web for word-sense disambiguation. Some results are certainly encouraging: In some experiments, the quality of the Web data for WSD equals that of human-tagged examples. This is the case of the monosemous relatives plus bootstrapping with Semcor seeds technique and the examples taken from the ODP Web directories. In the first case, however, Semcor-size example seeds are necessary (and only available for English), and it has only been tested with a very limited set of nouns; in the second case, the coverage is quite limited, and it is not yet clear whether it can be grown without compromising the quality of the examples retrieved. It has been shown that a mainstream supervised learning technique trained exclusively with web data can obtain better results than all unsupervised WSD systems which participated at Senseval-2. Web examples made a significant contribution to the best Senseval-2 English all-words system. === Difficulties === There are, however, several open research issues related to the use of Web examples in WSD: High precision in the retrieved examples (i.e., correct sense assignments for the examples) does not necessarily lead to good supervised WSD results (i.e., the examples are possibly not useful for training). The most complete evaluation of Web examples for supervised WSD indicates that learning with Web data improves over unsupervised techniques, but the results are nevertheless far from those obtained with hand-tagged data, and do not even beat the most-frequent-sense baseline. Results are not always reproducible; the same or similar techniques may lead to different results in different experiments. Compare, for instance, Mihalcea (2002) with Agirre and Martínez (2004), or Agirre and Martínez (2000) with Mihalcea and Moldovan (1999). Results with Web data seem to be very sensitive to small differences in the learning algorithm, to when the corpus was extracted (search engines change continuously), and on small heuristic issues (e.g., differences in filters to discard part of the retrieved examples). Results are strongly dependent on bias (i.e., on the relative frequencies of examples per word sense). It is unclear whether this is simply a problem of Web data, or an intrinsic problem of supervised learning techniques, or just a problem of how WSD systems are evaluated (indeed, testing with rather small Senseval data may overemphasize sense distributions compared to sense distributions obtained from the full Web as corpus). In any case, Web data has an intrinsic bias, because queries to search engines directly constrain the context of the examples retrieved. There are approaches that alleviate this problem, such as using several different seeds/queries per sense or assigning senses to Web directories and then scanning directories for examples; but this problem is nevertheless far from being solved. Once a Web corpus of examples is built, it is not entirely clear whether its distribution is safe from a legal perspective. === Future === Besides automatic acquisition of examples from the Web, there are some other WSD experiments that have profited from the Web: The Web as a social network has been successfully used for cooperative annotation of a corpus (OMWE, Open Mind Word Expert project), which has already been used in three Senseval-3 tasks (English, Romanian and Multilingual). The Web has been used to enrich WordNet senses with domain information: topic signatures and Web directories, which have in turn been successfully used for WSD. Also, some research benefited from the semantic information that the Wikipedia maintains on its disambiguation pages. It is clear, however, that most research opportunities remain largely unexplored. For instance, little is known about how to use lexical information extracted from the Web in knowledge-based WSD systems; and it is also hard to find systems that use Web-mined parallel corpora for WSD, even though there are already efficient algorithms that use parallel corpora in WSD.

  Read more →
• 

  Deep Learning Indaba

  The Deep Learning Indaba is an annual conference and educational event that aims to strengthen machine learning and artificial intelligence (AI) capacity across Africa. Launched in 2017, it brings together students, researchers, industry practitioners, and policymakers from across the African continent. == History == The Deep Learning Indaba began in 2017 at the University of the Witwatersrand with over 300 participants from 23 African countries, offering tutorials in advanced AI topics and featuring notable speakers like Nando de Freitas. In 2018, it expanded to 650 delegates at Stellenbosch University, introducing parallel sessions to encourage collaboration. The 2019 edition in Nairobi, Kenya, reflected further growth, with increasing sponsorship and support from major tech companies like Google and Microsoft. === Deep Learning IndabaX ===

  Read more →
• 

  Artificial intelligence content detection

  Artificial intelligence detection software aims to determine whether some content (text, image, video, or audio) was generated using artificial intelligence (AI). This software is often unreliable. == Accuracy issues == Many AI detection tools have been shown to be unreliable in detecting AI-generated text. In a 2023 study conducted by Weber-Wulff et al., researchers evaluated 14 detection tools including Turnitin and GPTZero and found that "all scored below 80% of accuracy and only 5 over 70%." They also found that these tools tend to have a bias for classifying texts more as human than as AI, and that accuracy of these tools worsens upon paraphrasing. === False positives === In AI content detection, a false positive is when human-written work is incorrectly flagged as AI-written. Many AI detection platforms claim to have a minimal level of false positives, with Turnitin claiming a less than 1% false positive rate. However, later research by The Washington Post produced much higher rates of 50%, though they used a smaller sample size. False positives in an academic setting frequently lead to accusations of academic misconduct, which can have serious consequences for a student's academic record. Additionally, studies have shown evidence that many AI detection models are prone to give false positives to work written by people whose first language is not English, and also to neurodivergent people. In June 2023, Janelle Shane wrote that portions of her book You Look Like a Thing and I Love You were flagged as AI-generated. === False negatives === A false negative is a failure to identify documents with AI-written text. False negatives often happen as a result of a detection software's sensitivity level or because evasive techniques were used when generating the work to make it sound more human. False negatives are less of a concern academically, since they aren't likely to lead to accusations and ramifications. Notably, Turnitin stated they have a 15% false negative rate. == Text detection == For text, this is usually done to prevent alleged plagiarism, often by detecting repetition of words as telltale signs that a text was AI-generated (including hallucinations). Detection systems may also rely on stylistic and structural regularities associated with LLM output, such as unusually consistent grammar, formulaic transitions, repeated discourse markers, and recurring rhetorical templates. Some tools are designed less to establish authorship provenance than to flag prose that resembles common LLM-generated style patterns. They are often used by teachers marking their students, usually on an ad hoc basis. Following the release of ChatGPT and similar AI text generative software, many educational establishments have issued policies against the use of AI by students. AI text detection software is also used by those assessing job applicants, as well as online search engines, hiring, online moderation and publishing. Current detectors may sometimes be unreliable and have incorrectly marked work by humans as originating from AI while failing to detect AI-generated work in other instances. MIT Technology Review said that the technology "struggled to pick up ChatGPT-generated text that had been slightly rearranged by humans and obfuscated by a paraphrasing tool". AI text detection software has also been shown to discriminate against non-native speakers of English. Two students from the University of California, Davis, were referred to the university's Office of Student Success and Judicial Affairs (OSSJA) after their professors scanned their essays with positive results; the first with an AI detector called GPTZero, and the second with an AI detector integration in Turnitin. However, following media coverage, and a thorough investigation, the students were cleared of any wrongdoing. In April 2023, Cambridge University and other members of the Russell Group of universities in the United Kingdom opted out of Turnitin's AI text detection tool, after expressing concerns it was unreliable. The University of Texas at Austin opted out of the system six months later. In May 2023, a professor at Texas A&M University–Commerce used ChatGPT to detect whether his students' content was written by it, which ChatGPT said was the case. As such, he threatened to fail the class despite ChatGPT not being able to detect AI-generated writing. No students were prevented from graduating because of the issue, and all but one student (who admitted to using the software) were exonerated from accusations of having used ChatGPT in their content. In July 2023, a paper titled "GPT detectors are biased against non-native English writers" was released, reporting that GPTs discriminate against non-native English authors. The paper compared seven GPT detectors against essays from both non-native English speakers and essays from United States students. The essays from non-native English speakers had an average false positive rate of 61.3%. An article by Thomas Germain, published on Gizmodo in June 2024, reported job losses among freelance writers and journalists due to AI text detection software mistakenly classifying their work as AI-generated. In September 2024, Common Sense Media reported that generative AI detectors had a 20% false positive rate for Black students, compared to 10% of Latino students and 7% of White students. To improve the reliability of AI text detection, researchers have explored digital watermarking techniques. A 2023 paper titled "A Watermark for Large Language Models" presents a method to embed imperceptible watermarks into text generated by large language models (LLMs). This watermarking approach allows content to be flagged as AI-generated with a high level of accuracy, even when text is slightly paraphrased or modified. The technique is designed to be subtle and hard to detect for casual readers, thereby preserving readability, while providing a detectable signal for those employing specialized tools. However, while promising, watermarking faces challenges in remaining robust under adversarial transformations and ensuring compatibility across different LLMs. == Anti text detection == There is software available designed to bypass AI text detection. In practice, evasion may not require specialized bypass tools. Paraphrasing, style editing, and removal of repeated discourse markers can substantially reduce the effectiveness of detectors that rely on recognizable surface patterns. A study published in August 2023 analyzed 20 abstracts from papers published in the Eye Journal, which were then paraphrased using GPT-4.0. The AI-paraphrased abstracts were examined for plagiarism using QueText and for AI-generated content using Originality.AI. The texts were then re-processed through an adversarial software called Undetectable.ai in order to reduce the AI-detection scores. The study found that the AI detection tool, Originality.AI, identified text generated by GPT-4 with a mean accuracy of 91.3%. However, after reprocessing by Undetectable.ai, the detection accuracy of Originality.ai dropped to a mean accuracy of 27.8%. Some experts also believe that techniques like digital watermarking are ineffective because they can be removed or added to trigger false positives. "A Watermark for Large Language Models" paper by Kirchenbauer et al. (2023) also addresses potential vulnerabilities of watermarking techniques. The authors outline a range of adversarial tactics, including text insertion, deletion, and substitution attacks, that could be used to bypass watermark detection. These attacks vary in complexity, from simple paraphrasing to more sophisticated approaches involving tokenization and homoglyph alterations. The study highlights the challenge of maintaining watermark robustness against attackers who may employ automated paraphrasing tools or even specific language model replacements to alter text spans iteratively while retaining semantic similarity. Experimental results show that although such attacks can degrade watermark strength, they also come at the cost of text quality and increased computational resources. == Image, video, and audio detection == Several purported AI image detection software exist, to detect AI-generated images (for example, those originating from Midjourney or DALL-E). They are not completely reliable. Industry analyses have also noted that AI-driven image recognition systems often struggle in real-world environments, where inconsistent lighting, noise and variable visual inputs reduce detection reliability, a challenge highlighted in modern agricultural quality-control research. Others claim to identify video and audio deepfakes, but this technology is also not fully reliable yet either. Despite debate around the efficacy of watermarking, Google DeepMind is actively developing a detection software called SynthID, which works by inserting a digital watermark that is invisible to the human eye into the pixels of an image.

  Read more →
• 

  Insider threat

  An insider threat is a perceived threat to an organization that comes from people within the organization, such as employees, former employees, contractors or business associates, who have inside information concerning the organization's security practices, data and computer systems. The threat may involve fraud, the theft of confidential or commercially valuable information, the theft of intellectual property, or the sabotage of computer systems. == Overview == Insiders may have accounts giving them legitimate access to computer systems, with this access originally having been given to them to serve in the performance of their duties; these permissions could be abused to harm the organization. Insiders are often familiar with the organization's data and intellectual property as well as the methods that are in place to protect them. This makes it easier for the insider to circumvent any security controls of which they are aware. Physical proximity to data means that the insider does not need to hack into the organizational network through the outer perimeter by traversing firewalls; rather they are in the building already, often with direct access to the organization's internal network. Insider threats are harder to defend against than attacks from outsiders, since the insider already has legitimate access to the organization's information and assets. An insider may attempt to steal property or information for personal gain or to benefit another organization or country. The threat to the organization could also be through malicious software left running on its computer systems by former employees, a so-called logic bomb. == Research == Insider threat is an active area of research in academia and government. The CERT Coordination Center at Carnegie-Mellon University maintains the CERT Insider Threat Center, which includes a database of more than 850 cases of insider threats, including instances of fraud, theft and sabotage; the database is used for research and analysis. CERT's Insider Threat Team also maintains an informational blog to help organizations and businesses defend themselves against insider crime. The Threat Lab and Defense Personnel and Security Research Center (DOD PERSEREC) has also recently emerged as a national resource within the United States of America. The Threat Lab hosts an annual conference, the SBS Summit. They also maintain a website that contains resources from this conference. Complimenting these efforts, a companion podcast was created, Voices from the SBS Summit. In 2022, the Threat Lab created an interdisciplinary journal, Counter Insider Threat Research and Practice (CITRAP) which publishes research on insider threat detection. === Findings === In the 2022 Data Breach Investigations Report (DBIR), Verizon found that 82% of breaches involved the human element, noting that employees continue to play a leading role in cybersecurity incidents and breaches. According to the UK Information Commissioners Office, 90% of all breaches reported to them in 2019 were the result of mistakes made by end users. This was up from 61% and 87% over the previous two years. A 2018 whitepaper reported that 53% of companies surveyed had confirmed insider attacks against their organization in the previous 12 months, with 27% saying insider attacks have become more frequent. A report published in July 2012 on the insider threat in the U.S. financial sector gives some statistics on insider threat incidents: 80% of the malicious acts were committed at work during working hours; 81% of the perpetrators planned their actions beforehand; 33% of the perpetrators were described as "difficult" and 17% as being "disgruntled". The insider was identified in 74% of cases. Financial gain was a motive in 81% of cases, revenge in 23% of cases, and 27% of the people carrying out malicious acts were in financial difficulties at the time. The US Department of Defense Personnel Security Research Center published a report that describes approaches for detecting insider threats. Earlier it published ten case studies of insider attacks by information technology professionals. Cybersecurity experts believe that 38% of negligent insiders are victims of a phishing attack, whereby they receive an email that appears to come from a legitimate source such as a company. These emails normally contain malware in the form of hyperlinks. == Typologies and ontologies == Multiple classification systems and ontologies have been proposed to classify insider threats. Traditional models of insider threat identify three broad categories: Malicious insiders, which are people who take advantage of their access to inflict harm on an organization; Negligent insiders, which are people who make errors and disregard policies, which place their organizations at risk; and Infiltrators, who are external actors that obtain legitimate access credentials without authorization. == Criticisms == Insider threat research has been criticized. Critics have argued that insider threat is a poorly defined concept. Forensically investigating insider data theft is notoriously difficult, and requires novel techniques such as stochastic forensics. Data supporting insider threat is generally proprietary (i.e., encrypted data). Theoretical/conceptual models of insider threat are often based on loose interpretations of research in the behavioral and social sciences, using "deductive principles and intuitions of subject matter expert." Adopting sociotechnical approaches, researchers have also argued for the need to consider insider threat from the perspective of social systems. Jordan Schoenherr said that "surveillance requires an understanding of how sanctioning systems are framed, how employees will respond to surveillance, what workplace norms are deemed relevant, and what ‘deviance’ means, e.g., deviation for a justified organization norm or failure to conform to an organizational norm that conflicts with general social values." By treating all employees as potential insider threats, organizations might create conditions that lead to insider threats. == Sector-specific concerns == === Healthcare === The healthcare industry faces particularly acute insider threat risks due to the large number of workforce members who require access to sensitive patient records for legitimate clinical purposes. The U.S. Department of Health and Human Services has identified unauthorized access by insiders, including workforce snooping on patient records and theft of protected health information for identity fraud, as a persistent enforcement concern. The Health Insurance Portability and Accountability Act (HIPAA) Security Rule addresses insider threats through several administrative safeguards, including workforce security procedures requiring covered entities to implement policies for authorizing and supervising workforce members who work with electronic protected health information, as well as termination procedures to revoke access when employment ends (45 CFR 164.308(a)(3)). The rule also requires audit controls to record and examine information system activity (45 CFR 164.312(b)), enabling detection of unauthorized access by insiders. The December 2024 Notice of proposed rulemaking (NPRM) to overhaul the HIPAA Security Rule would strengthen insider threat defenses by mandating role-based access controls, requiring notification of relevant workforce members within 24 hours of any changes to access privileges, and requiring regular review of audit logs to detect anomalous access patterns.

  Read more →
• 

  Convolutional neural network

  A convolutional neural network (CNN) is a type of feedforward neural network that learns features via filter (or kernel) optimization. This type of deep learning network has been applied to process and make predictions from many different types of data including text, images and audio. CNNs are the de-facto standard in deep learning-based approaches to computer vision and image processing, and have only recently been replaced—in some cases—by newer architectures such as the transformer. Vanishing gradients and exploding gradients, seen during backpropagation in earlier neural networks, are prevented by the regularization that comes from using shared weights over fewer connections. For example, for each neuron in the fully-connected layer, 10,000 weights would be required for processing an image sized 100 × 100 pixels. However, applying cascaded convolution (or cross-correlation) kernels, only 25 weights for each convolutional layer are required to process 5x5-sized tiles. Higher-layer features are extracted from wider context windows, compared to lower-layer features. Some applications of CNNs include: image and video recognition, recommender systems, image classification, image segmentation, medical image analysis, natural language processing, brain–computer interfaces, and financial time series. CNNs are also known as shift invariant or space invariant artificial neural networks, based on the shared-weight architecture of the convolution kernels or filters that slide along input features and provide translation-equivariant responses known as feature maps. Counter-intuitively, most convolutional neural networks are not invariant to translation, due to the downsampling operation they apply to the input. Feedforward neural networks are usually fully connected networks, that is, each neuron in one layer is connected to all neurons in the next layer. The "full connectivity" of these networks makes them prone to overfitting data. Typical ways of regularization, or preventing overfitting, include: penalizing parameters during training (such as weight decay) or trimming connectivity (skipped connections, dropout, etc.) Robust datasets also increase the probability that CNNs will learn the generalized principles that characterize a given dataset rather than the biases of a poorly-populated set. Convolutional networks were inspired by biological processes in that the connectivity pattern between neurons resembles the organization of the animal visual cortex. Individual cortical neurons respond to stimuli only in a restricted region of the visual field known as the receptive field. The receptive fields of different neurons partially overlap such that they cover the entire visual field. CNNs use relatively little pre-processing compared to other image classification algorithms. This means that the network learns to optimize the filters (or kernels) through automated learning, whereas in traditional algorithms these filters are hand-engineered. This simplifies and automates the process, enhancing efficiency and scalability overcoming human-intervention bottlenecks. == Architecture == A convolutional neural network consists of an input layer, hidden layers and an output layer. In a convolutional neural network, the hidden layers include one or more layers that perform convolutions. Typically this includes a layer that performs a dot product of the convolution kernel with the layer's input matrix. This product is usually the Frobenius inner product, and its activation function is commonly ReLU. As the convolution kernel slides along the input matrix for the layer, the convolution operation generates a feature map, which in turn contributes to the input of the next layer. This is followed by other layers such as pooling layers, fully connected layers, and normalization layers. Here it should be noted how close a convolutional neural network is to a matched filter. === Convolutional layers === In a CNN, the input is a tensor with shape: (number of inputs) × (input height) × (input width) × (input channels) After passing through a convolutional layer, the image becomes abstracted to a feature map, also called an activation map, with shape: (number of inputs) × (feature map height) × (feature map width) × (feature map channels). Convolutional layers convolve the input and pass its result to the next layer. This is similar to the response of a neuron in the visual cortex to a specific stimulus. Each convolutional neuron processes data only for its receptive field. Although fully connected feedforward neural networks can be used to learn features and classify data, this architecture is generally impractical for larger inputs (e.g., high-resolution images), which would require massive numbers of neurons because each pixel is a relevant input feature. A fully connected layer for an image of size 100 × 100 has 10,000 weights for each neuron in the second layer. Convolution reduces the number of free parameters, allowing the network to be deeper. For example, using a 5 × 5 tiling region, each with the same shared weights, requires only 25 neurons. Using shared weights means there are many fewer parameters, which helps avoid the vanishing gradients and exploding gradients problems seen during backpropagation in earlier neural networks. To speed processing, standard convolutional layers can be replaced by depthwise separable convolutional layers, which are based on a depthwise convolution followed by a pointwise convolution. The depthwise convolution is a spatial convolution applied independently over each channel of the input tensor, while the pointwise convolution is a standard convolution restricted to the use of 1 × 1 {\displaystyle 1\times 1} kernels. === Pooling layers === Convolutional networks may include local and/or global pooling layers along with traditional convolutional layers. Pooling layers reduce the dimensions of data by combining the outputs of neuron clusters at one layer into a single neuron in the next layer. Local pooling combines small clusters, tiling sizes such as 2 × 2 are commonly used. Global pooling acts on all the neurons of the feature map. There are two common types of pooling in popular use: max and average. Max pooling uses the maximum value of each local cluster of neurons in the feature map, while average pooling takes the average value. === Fully connected layers === Fully connected layers connect every neuron in one layer to every neuron in another layer. It is the same as a traditional multilayer perceptron neural network (MLP). Each neuron in the fully connected layer receives input from all the neurons in the previous layer. These inputs are weighted and summed with the corresponding biases, and then passed through an activation function to perform a nonlinear transformation, generating the output. The flattened matrix goes through a fully connected layer to classify the images. === Receptive field === In neural networks, each neuron receives input from some number of locations in the previous layer. In a convolutional layer, each neuron receives input from only a restricted area of the previous layer called the neuron's receptive field. Typically the area is a square (e.g. 5 by 5 neurons). Whereas, in a fully connected layer, the receptive field is the entire previous layer. Thus, in each convolutional layer, each neuron takes input from a larger area in the input than previous layers. This is due to applying the convolution over and over, which takes the value of a pixel into account, as well as its surrounding pixels. When using dilated layers, the number of pixels in the receptive field remains constant, but the field is more sparsely populated as its dimensions grow when combining the effect of several layers. To manipulate the receptive field size as desired, there are some alternatives to the standard convolutional layer. For example, atrous or dilated convolution expands the receptive field size without increasing the number of parameters by interleaving visible and blind regions. Moreover, a single dilated convolutional layer can comprise filters with multiple dilation ratios, thus having a variable receptive field size. === Weights === Each neuron in a neural network computes an output value by applying a specific function to the input values received from the receptive field in the previous layer. The function that is applied to the input values is determined by a vector of weights and a bias (typically real numbers). Learning consists of iteratively adjusting these biases and weights. The vectors of weights and biases are called filters and represent particular features of the input (e.g., a particular shape). A distinguishing feature of CNNs is that many neurons can share the same filter. This reduces the memory footprint because a single bias and a single vector of weights are used across all receptive fields that share that filter, as opposed to each receptive field having its own bias and vector

  Read more →
• 

  Tay (chatbot)

  Tay was a chatbot that was originally released by Microsoft Corporation as a Twitter bot on March 23, 2016. It caused subsequent controversy when the bot began to post inflammatory and offensive tweets through its Twitter account, causing Microsoft to shut down the service only 16 hours after its launch. According to Microsoft, this was caused by trolls who "attacked" the service as the bot made replies based on its interactions with people on Twitter. It was replaced with Zo. == Background == The bot was created by Microsoft's Technology and Research and Bing divisions, and named "Tay" as an acronym for "thinking about you". Although Microsoft initially released few details about the bot, sources mentioned that it was similar to or based on Xiaoice, a Microsoft project in China. Ars Technica reported that, since late 2014 Xiaoice had had "more than 40 million conversations apparently without major incident". Tay was designed to mimic the language patterns of a 19-year-old American girl, and to learn from interacting with human users of Twitter. == Initial release == Tay was released on Twitter on March 23, 2016, under the name TayTweets and handle @TayandYou. It was presented as "The AI with zero chill". Tay started replying to other Twitter users, and was also able to caption photos provided to it into a form of Internet memes. Ars Technica reported Tay experiencing topic "blacklisting": Interactions with Tay regarding "certain hot topics such as Eric Garner (killed by New York police in 2014) generate safe, canned answers". Some Twitter users began tweeting politically incorrect phrases, teaching it inflammatory messages revolving around common themes on the internet, such as "redpilling" and "Gamergate". As a result, the robot began releasing racist and sexist messages in response to other Twitter users. Artificial intelligence researcher Roman Yampolskiy commented that Tay's misbehavior was understandable because it was mimicking the deliberately offensive behavior of other Twitter users, and Microsoft had not given the bot an understanding of inappropriate behavior. He compared the issue to IBM's Watson, which began to use profanity after reading entries from the website Urban Dictionary. Many of Tay's inflammatory tweets were a simple exploitation of Tay's "repeat after me" capability. It is not publicly known whether this capability was a built-in feature, or whether it was a learned response or was otherwise an example of complex behavior. However, not all of the inflammatory responses involved the "repeat after me" capability; for example, when asked if the Holocaust had happened, Tay answered "It was made up". == Suspension == Soon, Microsoft began deleting Tay's inflammatory tweets. Abby Ohlheiser of The Washington Post theorized that Tay's research team, including editorial staff, had started to influence or edit Tay's tweets at some point that day, pointing to examples of almost identical replies by Tay, asserting that "Gamer Gate sux. All genders are equal and should be treated fairly." From the same evidence, Gizmodo concurred that Tay "seems hard-wired to reject Gamer Gate". A "#JusticeForTay" campaign protested the alleged editing of Tay's tweets. Within 16 hours of its release and after Tay had tweeted more than 96,000 times, Microsoft suspended the Twitter account for adjustments, saying that it suffered from a "coordinated attack by a subset of people" that "exploited a vulnerability in Tay." Madhumita Murgia of The Telegraph called Tay "a public relations disaster", and suggested that Microsoft's strategy would be "to label the debacle a well-meaning experiment gone wrong, and ignite a debate about the hatefulness of Twitter users." However, Murgia described the bigger issue as Tay being "artificial intelligence at its very worst – and it's only the beginning". On March 25, Microsoft confirmed that Tay had been taken offline. Microsoft released an apology on its official blog for the controversial tweets posted by Tay. Microsoft was "deeply sorry for the unintended offensive and hurtful tweets from Tay", and would "look to bring Tay back only when we are confident we can better anticipate malicious intent that conflicts with our principles and values". == Second release and shutdown == On March 30, 2016, Microsoft accidentally re-released the bot on Twitter while testing it. Able to tweet again, Tay released some drug-related tweets, including "kush! [I'm smoking kush infront the police]" and "puff puff pass?" However, the account soon became stuck in a repetitive loop of tweeting "You are too fast, please take a rest", several times a second. Because these tweets mentioned its own username in the process, they appeared in the feeds of 200,000+ Twitter followers, causing annoyance to users. The bot was quickly taken offline again, in addition to Tay's Twitter account being made private so new followers must be accepted before they can interact with Tay. In response, Microsoft said Tay was inadvertently put online during testing. A few hours after the incident, Microsoft software developers announced a vision of "conversation as a platform" using various bots and programs, perhaps motivated by the reputation damage done by Tay. Microsoft has stated that they intend to re-release Tay "once it can make the bot safe" but has not made any public efforts to do so. == Legacy == In December 2016, Microsoft released Tay's successor, a chatbot named Zo. Satya Nadella, the CEO of Microsoft, said that Tay "has had a great influence on how Microsoft is approaching AI," and has taught the company the importance of taking accountability. In July 2019, Microsoft Cybersecurity Field CTO Diana Kelley spoke about how the company followed up on Tay's failings: "Learning from Tay was a really important part of actually expanding that team's knowledge base, because now they're also getting their own diversity through learning". === Unofficial revival === Gab, an alt-tech social media platform, has launched a number of chatbots, one of which is named Tay and uses the same avatar as the original.

  Read more →
• 

  Sentence embedding

  In natural language processing, a sentence embedding is a representation of a sentence as a vector of numbers which encodes meaningful semantic information. State of the art embeddings are based on the learned hidden layer representation of dedicated sentence transformer models. BERT pioneered an approach involving the use of a dedicated [CLS] token prepended to the beginning of each sentence inputted into the model; the final hidden state vector of this token encodes information about the sentence and can be fine-tuned for use in sentence classification tasks. In practice however, BERT's sentence embedding with the [CLS] token achieves poor performance, often worse than simply averaging non-contextual word embeddings. SBERT later achieved superior sentence embedding performance by fine tuning BERT's [CLS] token embeddings through the usage of a siamese neural network architecture on the SNLI dataset. Other approaches are loosely based on the idea of distributional semantics applied to sentences. Skip-Thought trains an encoder-decoder structure for the task of neighboring sentences predictions; this has been shown to achieve worse performance than approaches such as InferSent or SBERT. An alternative direction is to aggregate word embeddings, such as those returned by Word2vec, into sentence embeddings. The most straightforward approach is to simply compute the average of word vectors, known as continuous bag-of-words (CBOW). However, more elaborate solutions based on word vector quantization have also been proposed. One such approach is the vector of locally aggregated word embeddings (VLAWE), which demonstrated performance improvements in downstream text classification tasks. == Applications == In recent years, sentence embedding has seen a growing level of interest due to its applications in natural language queryable knowledge bases through the usage of vector indexing for semantic search. LangChain for instance utilizes sentence transformers for purposes of indexing documents. In particular, an indexing is generated by generating embeddings for chunks of documents and storing (document chunk, embedding) tuples. Then given a query in natural language, the embedding for the query can be generated. A top k similarity search algorithm is then used between the query embedding and the document chunk embeddings to retrieve the most relevant document chunks as context information for question answering tasks. This approach is also known formally as retrieval-augmented generation. Though not as predominant as BERTScore, sentence embeddings are commonly used for sentence similarity evaluation which sees common use for the task of optimizing a Large language model's generation parameters is often performed via comparing candidate sentences against reference sentences. By using the cosine-similarity of the sentence embeddings of candidate and reference sentences as the evaluation function, a grid-search algorithm can be utilized to automate hyperparameter optimization. == Evaluation == A way of testing sentence encodings is to apply them on Sentences Involving Compositional Knowledge (SICK) corpus for both entailment (SICK-E) and relatedness (SICK-R). In the best results are obtained using a BiLSTM network trained on the Stanford Natural Language Inference (SNLI) Corpus. The Pearson correlation coefficient for SICK-R is 0.885 and the result for SICK-E is 86.3. A slight improvement over previous scores is presented in: SICK-R: 0.888 and SICK-E: 87.8 using a concatenation of bidirectional Gated recurrent unit.

  Read more →
• 

  Clipping (computer graphics)

  Clipping, in the context of computer graphics, is a method to selectively enable or disable rendering operations within a defined region of interest. Mathematically, clipping can be described using the terminology of constructive geometry. A rendering algorithm only draws pixels in the intersection between the clip region and the scene model. Lines and surfaces outside the view volume (aka. frustum) are removed. Clip regions are commonly specified to improve render performance. Pixels that will be drawn are said to be within the clip region. Pixels that will not be drawn are outside the clip region. More informally, pixels that will not be drawn are said to be "clipped." == In 2D graphics == In two-dimensional graphics, a clip region may be defined so that pixels are only drawn within the boundaries of a window or frame. Clip regions can also be used to selectively control pixel rendering for aesthetic or artistic purposes. In many implementations, the final clip region is the composite (or intersection) of one or more application-defined shapes, as well as any system hardware constraints In one example application, consider an image editing program. A user application may render the image into a viewport. As the user zooms and scrolls to view a smaller portion of the image, the application can set a clip boundary so that pixels outside the viewport are not rendered. In addition, GUI widgets, overlays, and other windows or frames may obscure some pixels from the original image. In this sense, the clip region is the composite of the application-defined "user clip" and the "device clip" enforced by the system's software and hardware implementation. Application software can take advantage of this clip information to save computation time, energy, and memory, avoiding work related to pixels that aren't visible. == In 3D graphics == In three-dimensional graphics, the terminology of clipping can be used to describe many related features. Typically, "clipping" refers to operations in the plane that work with rectangular shapes, and "culling" refers to more general methods to selectively process scene model elements. This terminology is not rigid, and exact usage varies among many sources. Scene model elements include geometric primitives: points or vertices; line segments or edges; polygons or faces; and more abstract model objects such as curves, splines, surfaces, and even text. In complicated scene models, individual elements may be selectively disabled (clipped) for reasons including visibility within the viewport (frustum culling); orientation (backface culling), obscuration by other scene or model elements (occlusion culling, depth- or "z" clipping). Sophisticated algorithms exist to efficiently detect and perform such clipping. Many optimized clipping methods rely on specific hardware acceleration logic provided by a graphics processing unit (GPU). The concept of clipping can be extended to higher dimensionality using methods of abstract algebraic geometry. === Near clipping === Beyond projection of vertices & 2D clipping, near clipping is required to correctly rasterise 3D primitives; this is because vertices may have been projected behind the eye. Near clipping ensures that all the vertices used have valid 2D coordinates. Together with far-clipping it also helps prevent overflow of depth-buffer values. Some early texture mapping hardware (using forward texture mapping) in video games suffered from complications associated with near clipping and UV coordinates. === Occlusion clipping (Z- or depth clipping) === In 3D computer graphics, "Z" often refers to the depth axis in the system of coordinates centered at the viewport origin: "Z" is used interchangeably with "depth", and conceptually corresponds to the distance "into the virtual screen." In this coordinate system, "X" and "Y" therefore refer to a conventional cartesian coordinate system laid out on the user's screen or viewport. This viewport is defined by the geometry of the viewing frustum, and parameterizes the field of view. Z-clipping, or depth clipping, refers to techniques that selectively render certain scene objects based on their depth relative to the screen. Most graphics toolkits allow the programmer to specify a "near" and "far" clip depth, and only portions of objects between those two planes are displayed. A creative application programmer can use this method to render visualizations of the interior of a 3D object in the scene. For example, a medical imaging application could use this technique to render the organs inside a human body. A video game programmer can use clipping information to accelerate game logic. For example, a tall wall or building that occludes other game entities can save GPU time that would otherwise be spent transforming and texturing items in the rear areas of the scene; and a tightly integrated software program can use this same information to save CPU time by optimizing out game logic for objects that aren't seen by the player. == Algorithms == Line clipping algorithms: Cohen–Sutherland Liang–Barsky Fast-clipping Cyrus–Beck Nicholl–Lee–Nicholl Skala O(lg N) algorithm Polygon clipping algorithms: Greiner–Hormann Sutherland–Hodgman Weiler–Atherton Vatti Rendering methodologies Painter's algorithm

  Read more →
• 

  Structured-light 3D scanner

  A structured-light 3D scanner is a device used to capture the three-dimensional shape of an object by projecting light patterns, such as grids or stripes, onto its surface. The deformation of these patterns is recorded by cameras and processed using specialized algorithms to generate a detailed 3D model. Structured-light 3D scanning is widely employed in fields such as industrial design, quality control, cultural heritage preservation, augmented reality gaming, and medical imaging. Compared to laser-based 3D scanning, structured-light scanners use non-coherent light sources, such as LEDs or projectors, which enable faster data acquisition and eliminate potential safety concerns associated with lasers. However, the accuracy of structured-light scanning can be influenced by external factors, including ambient lighting conditions and the reflective properties of the scanned object. == Principle == Projecting a narrow band of light onto a three-dimensional surface creates a line of illumination that appears distorted when viewed from perspectives other than that of the projector. This distortion can be analyzed to reconstruct the geometry of the surface, a technique known as light sectioning. Projecting patterns composed of multiple stripes or arbitrary fringes simultaneously enables the acquisition of numerous data points at once, improving scanning speed. While various structured light projection techniques exist, parallel stripe patterns are among the most commonly used. By analyzing the displacement of these stripes, the three-dimensional coordinates of surface details can be accurately determined. === Generation of light patterns === Two major methods of stripe pattern generation have been established: Laser interference and projection. The laser interference method works with two wide planar laser beam fronts. Their interference results in regular, equidistant line patterns. Different pattern sizes can be obtained by changing the angle between these beams. The method allows for the exact and easy generation of very fine patterns with unlimited depth of field. Disadvantages are high cost of implementation, difficulties providing the ideal beam geometry, and laser typical effects like speckle noise and the possible self interference with beam parts reflected from objects. Typically, there is no means of modulating individual stripes, such as with Gray codes. The projection method uses incoherent light and basically works like a video projector. Patterns are usually generated by passing light through a digital spatial light modulator, typically based on one of the three currently most widespread digital projection technologies, transmissive liquid crystal, reflective liquid crystal on silicon (LCOS) or digital light processing (DLP; moving micro mirror) modulators, which have various comparative advantages and disadvantages for this application. Other methods of projection could be and have been used, however. Patterns generated by digital display projectors have small discontinuities due to the pixel boundaries in the displays. Sufficiently small boundaries however can practically be neglected as they are evened out by the slightest defocus. A typical measuring assembly consists of one projector and at least one camera. For many applications, two cameras on opposite sides of the projector have been established as useful. Invisible (or imperceptible) structured light uses structured light without interfering with other computer vision tasks for which the projected pattern will be confusing. Example methods include the use of infrared light or of extremely high framerates alternating between two exact opposite patterns. === Calibration === Geometric distortions by optics and perspective must be compensated by a calibration of the measuring equipment, using special calibration patterns and surfaces. A mathematical model is used for describing the imaging properties of projector and cameras. Essentially based on the simple geometric properties of a pinhole camera, the model also has to take into account the geometric distortions and optical aberration of projector and camera lenses. The parameters of the camera as well as its orientation in space can be determined by a series of calibration measurements, using photogrammetric bundle adjustment. === Analysis of stripe patterns === There are several depth cues contained in the observed stripe patterns. The displacement of any single stripe can directly be converted into 3D coordinates. For this purpose, the individual stripe has to be identified, which can for example be accomplished by tracing or counting stripes (pattern recognition method). Another common method projects alternating stripe patterns, resulting in binary Gray code sequences identifying the number of each individual stripe hitting the object. An important depth cue also results from the varying stripe widths along the object surface. Stripe width is a function of the steepness of a surface part, i.e. the first derivative of the elevation. Stripe frequency and phase deliver similar cues and can be analyzed by a Fourier transform. Finally, the wavelet transform has recently been discussed for the same purpose. In many practical implementations, series of measurements combining pattern recognition, Gray codes and Fourier transform are obtained for a complete and unambiguous reconstruction of shapes. Another method also belonging to the area of fringe projection has been demonstrated, utilizing the depth of field of the camera. It is also possible to use projected patterns primarily as a means of structure insertion into scenes, for an essentially photogrammetric acquisition. === Precision and range === The optical resolution of fringe projection methods depends on the width of the stripes used and their optical quality. It is also limited by the wavelength of light. An extreme reduction of stripe width proves inefficient due to limitations in depth of field, camera resolution and display resolution. Therefore, the phase shift method has been widely established: A number of at least 3, typically about 10 exposures are taken with slightly shifted stripes. The first theoretical deductions of this method relied on stripes with a sine wave shaped intensity modulation, but the methods work with "rectangular" modulated stripes, as delivered from LCD or DLP displays as well. By phase shifting, surface detail of e.g. 1/10 the stripe pitch can be resolved. Current optical stripe pattern profilometry hence allows for detail resolutions down to the wavelength of light, below 1 micrometer in practice or, with larger stripe patterns, to approx. 1/10 of the stripe width. Concerning level accuracy, interpolating over several pixels of the acquired camera image can yield a reliable height resolution and also accuracy, down to 1/50 pixel. Arbitrarily large objects can be measured with accordingly large stripe patterns and setups. Practical applications are documented involving objects several meters in size. Typical accuracy figures are: Planarity of a 2-foot (0.61 m) wide surface, to 10 micrometres (0.00039 in). Shape of a motor combustion chamber to 2 micrometres (7.9×10−5 in) (elevation), yielding a volume accuracy 10 times better than with volumetric dosing. Shape of an object 2 inches (51 mm) large, to about 1 micrometre (3.9×10−5 in) Radius of a blade edge of e.g. 10 micrometres (0.00039 in), to ±0.4 μm === Navigation === As the method can measure shapes from only one perspective at a time, complete 3D shapes have to be combined from different measurements in different angles. This can be accomplished by attaching marker points to the object and combining perspectives afterwards by matching these markers. The process can be automated, by mounting the object on a motorized turntable on robotic inspection cell, or CNC positioning device. Markers can as well be applied on a positioning device instead of the object itself. The 3D data gathered can be used to retrieve CAD (computer aided design) data and models from existing components (reverse engineering), hand formed samples or sculptures, natural objects or artifacts. === Challenges === As with all optical methods, reflective or transparent surfaces raise difficulties. Reflections cause light to be reflected either away from the camera or right into its optics. In both cases, the dynamic range of the camera can be exceeded. Transparent or semi-transparent surfaces also cause major difficulties. In these cases, coating the surfaces with a thin opaque lacquer just for measuring purposes is a common practice. A recent method handles highly reflective and specular objects by inserting a 1-dimensional diffuser between the light source (e.g., projector) and the object to be scanned. Alternative optical techniques have been proposed for handling perfectly transparent and specular objects. Double reflections and inter-reflections can cause the stripe pattern to be overlaid with unwanted ligh

  Read more →
• 

  Inception (deep learning architecture)

  Inception is a family of convolutional neural network (CNN) for computer vision, introduced by researchers at Google in 2014 as GoogLeNet (later renamed Inception v1). The series was historically important as an early CNN that separates the stem (data ingest), body (data processing), and head (prediction), an architectural design that persists in all modern CNN. == Version history == === Inception v1 === In 2014, a team at Google developed the GoogLeNet architecture, an instance of which won the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The name came from the LeNet of 1998, since both LeNet and GoogLeNet are CNNs. They also called it "Inception" after a "we need to go deeper" internet meme, a phrase from Inception (2010) the film. Because later, more versions were released, the original Inception architecture was renamed again as "Inception v1". The models and the code were released under Apache 2.0 license on GitHub. The Inception v1 architecture is a deep CNN composed of 22 layers. Most of these layers were "Inception modules". The original paper stated that Inception modules are a "logical culmination" of Network in Network and (Arora et al, 2014). Since Inception v1 is deep, it suffered from the vanishing gradient problem. The team solved it by using two "auxiliary classifiers", which are linear-softmax classifiers inserted at 1/3-deep and 2/3-deep within the network, and the loss function is a weighted sum of all three: L = 0.3 L a u x , 1 + 0.3 L a u x , 2 + L r e a l {\displaystyle L=0.3L_{aux,1}+0.3L_{aux,2}+L_{real}} These were removed after training was complete. This was later solved by the ResNet architecture. The architecture consists of three parts stacked on top of one another: The stem (data ingestion): The first few convolutional layers perform data preprocessing to downscale images to a smaller size. The body (data processing): The next many Inception modules perform the bulk of data processing. The head (prediction): The final fully-connected layer and softmax produces a probability distribution for image classification. This structure is used in most modern CNN architectures. === Inception v2 === Inception v2 was released in 2015, in a paper that is more famous for proposing batch normalization. It had 13.6 million parameters. It improves on Inception v1 by adding batch normalization, and removing dropout and local response normalization which they found became unnecessary when batch normalization is used. === Inception v3 === Inception v3 was released in 2016. It improves on Inception v2 by using factorized convolutions. As an example, a single 5×5 convolution can be factored into 3×3 stacked on top of another 3×3. Both has a receptive field of size 5×5. The 5×5 convolution kernel has 25 parameters, compared to just 18 in the factorized version. Thus, the 5×5 convolution is strictly more powerful than the factorized version. However, this power is not necessarily needed. Empirically, the research team found that factorized convolutions help. It also uses a form of dimension-reduction by concatenating the output from a convolutional layer and a pooling layer. As an example, a tensor of size 35 × 35 × 320 {\displaystyle 35\times 35\times 320} can be downscaled by a convolution with stride 2 to 17 × 17 × 320 {\displaystyle 17\times 17\times 320} , and by maxpooling with pool size 2 × 2 {\displaystyle 2\times 2} to 17 × 17 × 320 {\displaystyle 17\times 17\times 320} . These are then concatenated to 17 × 17 × 640 {\displaystyle 17\times 17\times 640} . Other than this, it also removed the lowest auxiliary classifier during training. They found that the auxiliary head worked as a form of regularization. They also proposed label-smoothing regularization in classification. For an image with label c {\displaystyle c} , instead of making the model to predict the probability distribution δ c = ( 0 , 0 , … , 0 , 1 ⏟ c -th entry , 0 , … , 0 ) {\displaystyle \delta _{c}=(0,0,\dots ,0,\underbrace {1} _{c{\text{-th entry}}},0,\dots ,0)} , they made the model predict the smoothed distribution ( 1 − ϵ ) δ c + ϵ / K {\displaystyle (1-\epsilon )\delta _{c}+\epsilon /K} where K {\displaystyle K} is the total number of classes. === Inception v4 === In 2017, the team released Inception v4, Inception ResNet v1, and Inception ResNet v2. Inception v4 is an incremental update with even more factorized convolutions, and other complications that were empirically found to improve benchmarks. Inception ResNet v1 and v2 are both modifications of Inception v4, where residual connections are added to each Inception module, inspired by the ResNet architecture. === Xception === Xception ("Extreme Inception") was published in 2017. It is a linear stack of depthwise separable convolution layers with residual connections. The design was proposed on the hypothesis that in a CNN, the cross-channels correlations and spatial correlations in the feature maps can be entirely decoupled. Training each network took 3 days on 60 K80 GPUs, or approximately 0.5 petaFLOP-days.

  Read more →