BeReal (stylized on the app logo as BeReal.) is a French social-networking app released in 2020, developed by Alexis Barreyat and Kévin Perreau. Currently, it is owned by Voodoo. Its main feature is a daily notification that encourages users to share photos of themselves in their day-to-day life, on any randomly selected two-minute window every day. Critics noted its emphasis on authenticity, which some felt crossed the line into the mundane. The primary reference of its name relates to its focus on users uploading unpolished photos, with it being a pun of the term B-reel. According to the app's description on Apple's App Store, BeReal encourages its users to "show their friends who they really are, for once," by removing filters and opportunities to stage or edit photos. After a couple of years of relative obscurity, it rapidly gained popularity in early and mid-2022 growing from 21.6 million to 73.5 million users between July and August, before experiencing a decrease in use in 2023 and continuing to decline to 23 million users at the beginning of 2024. == History == The app was developed by Alexis Barreyat, a former employee at GoPro, and Kévin Perreau, a graduate from 42 in Paris. Initially released in 2020, it first gained widespread popularity in early 2022. It first spread widely on college campuses, partially due to a paid ambassador program. In late August 2022, the application had over 10 million active daily users and 21.6 million active monthly users. As of February 2023, the app has grown to 13 million active daily users and 47.8 million active monthly users. In June 2021, BeReal received a $30 million funding round led by Andreessen Horowitz and Accel. In May 2022, BeReal secured $85 million in a funding round led by Yuri Milner's DST Global, increasing its valuation to about $600 million. On July 25, 2022, BeReal topped Apple's free app list in the iOS App Store, and remained until September 2022. BeReal also received Apple's iPhone App of the Year in 2022. By late spring 2023, the app's momentum was waning, as daily users dropped to about 6 million, from 15 million in October 2022. In August 2024, there was a resurgence after a campaign at the Paris Olympics 2024, with the app reportedly gaining 1000 users. In June 2024, BeReal was acquired by the French company Voodoo for a reported €500 million. Alexis Barreyat is set to step down after a transition period. == Features == Once per day, BeReal notifies all users that a two-minute window to post is open. It asks users to create a post (known eponymously as a "BeReal") which, using mandatory simultaneous photos and now short videos from both the front and back cameras, provides a visual depiction of what they are doing at that moment, with an option to caption their post. The given window varies from day to day, and is not known to users before the notification is received. Once the daily notification is sent, users lose the ability to see others' BeReals from the previous day. Furthermore, users cannot see any of the current day's BeReals until they upload their own. On-time BeReals show the time it was uploaded, meanwhile, late BeReals uploaded after the two-minute window shows how late the BeReal was taken, but the user has to long-press the BeReal to reveal the time it was uploaded. Other users can also see how many attempts the poster took to take the BeReal, as well as their location when the BeReal was taken. Users only get one chance to delete their BeReal and post another one, and they used to not be able to post more than one at any time. However, in 2023, a feature was added that allowed users to post up to two extra BeReals on days when they posted their first BeReal within the 2-minute window. In July 2024, the number of bonus BeReals was increased to 5. [1] BeReal also features a "Discovery" section, wherein users are given the option to share to a much wider, public audience. This feature, however, is limited, as users are not able to interact with the posts through commenting—unlike the "My Friends" feature. In August 2023, in an attempt to make BeReal more social, another feature was added so that users are now able to see their friends of friends' BeReal. The app reportedly uses HiveAI to automate its image moderation process. However, there is also a report function that allows users to report a photo or another user if they are posting inappropriate content. === Comparison to other platforms === Because of its daily cycle of engagement, it has been compared to Wordle, which gained popularity earlier in 2022. It also supports a platform similar to Snapchat with a theme of impermanence and brevity. BeReal has been described as designed to compete with Instagram while simultaneously de-emphasising social media addiction and overuse. The app does not allow any photo filters or other editing, and has no follower counts. Marketing material from the company said that the app "can be addictive" and that "BeReal won't make you famous." Jacob Arnott, managing director of social agency We the People, describes BeReal as "an anti-Instagram" due to its raw and unedited nature. The app's foundation on friends rather than followers resembles Facebook's platform of adding friends, which comprise the content of a user's feed. This also resembles Instagram's "close friends" story feature. Further, rather than "liking" posts, BeReal uses "RealMojis" which involves taking a photo to interact with other posts. With the popularity of BeReal, other providers have launched similar features. In July 2022, Instagram launched a "Dual Camera" feature similar to BeReal, and in August 2022 it began testing a feature called "IG Candid Challenges", where users are prompted to post once a day within two minutes. As of September 2022, TikTok has also launched a feature called TikTok Now, following the same concept. In December 2022, similar to Spotify's "Wrapped," BeReal launched a feature involving a video of a compilation of users' BeReal posts of 2022. == User characteristics == BeReal is considered to be targeted towards Generation Z users, and attempts to minimise "social media fatigue", a feeling of numbness and disconnection from reality caused by constant interaction with an idealised version of others. This is a "core generational value" that this demographic holds compared to Millennials. Further, BeReal's users have been particularly strong across universities and university-aged students, and the majority of users are in the United States, the United Kingdom, and Germany. In 2022, the majority of users were female, with 43.2% of users falling within the age range of 16 to 25 and 55.1% of users being 26 to 44 years old. BeReal, the platform encourages users to share their real time moments by sending a daily notification that gives a least two minutes to post a unedited photo using bot the front and back camera, although users can post later and retake photos from when the notification happens, this action are still visible to friends, reinforcing transparency and genuine in the moment sharing. == Reception == Jason Koebler, a writer for Vice, wrote that in contrast to Instagram, which presents an unattainable view of people's lives, BeReal instead "makes everyone look extremely boring". Niklas Myhr, a professor of social media at Chapman University, argued that depth of engagement may determine whether the app is a passing trend or has "staying power". Kelsey Weekman, a reporter for BuzzFeed News, noted that the app's unwillingness to "glamorise the banality of life" made it feel "humbling" in its emphasis on authenticity. Niloufar Haidari for The Guardian comments similarly that where the app succeeds in being "drab" in perhaps a positive way, it fails in potentially "un-inspiring" users. Likewise, Dr. Brad Ridout, a behavioral psychologist at the University of Sydney, emphasizes that the "boring" experience is what the creators are targeting for the app and, in response to Instagram's platform of flawlessness, that "perfection is the enemy of happiness". === Criticisms === Some people regularly post after the two-minute notification expires, leading to some criticism of the app, as the ability to post late undermines its aims of authenticity. In addition, BeReal's daily two-minute window has been argued to contribute to social media fatigue and a need for self-exposure, as well as constant access to phones.
KeyBase
KeyBase is a database and web application for managing and deploying interactive taxonomic keys for plants and animals developed by the Royal Botanic Gardens Victoria. KeyBase provides a medium where pathway keys which were traditionally developed for print and other classical types of media, can be used more effectively in the internet environment. The platform uses a concept called "keys" which can be easily linked together, joined with other keys, or merged into larger other seamless keys groups, with each still available to be browsed independently. Keys in the KeyBase database can be filtered and displayed in a variety of ways, filters, and formats.
International Conference on Autonomous Agents and Multiagent Systems
The International Conference on Autonomous Agents and Multiagent Systems or AAMAS is the leading scientific conference for research in the areas of artificial intelligence, autonomous agents, and multiagent systems. It is annually organized by a non-profit organization called the International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS). == History == The International Conference on Autonomous Agents and Multiagent Systems (AAMAS) is a highly respected joint conference that provides a quality forum for discussing research in intelligent computational agents and their interactions. It is a merger of three major international conferences/workshops, namely the International Conference on Autonomous Agents (AGENTS), International Conference on Multi-Agent Systems (ICMAS), and International Workshop on Agent Theories, Architectures, and Languages (ATAL). ICMAS is itself a merger of three formative workshops, each with an attendance of fewer than 50 researchers. At a meeting during IJCAI-93 held in Chambery, France in August 1993, the leaders of the European Workshops on Modelling Autonomous Agents in a Multi-Agent World, the Asian MAAC Workshops, and the North American Distributed Artificial Intelligence Workshops (Victor Lesser, Michael N. Huhns, Les Gasser, Barbara Grosz, Nicholas Jennings, Michael Wooldridge, Gerhard Weiss, Mario Tokoro, and Toru Ishida) began the planning for a combined conference, which resulted in the first ICMAS in San Francisco, CA, USA in 1995, attended by more than 500 researchers. The AAMAS Conference is under the guidance and management of the International Foundation for Autonomous Agents and Multiagent Systems, which is incorporated as a 501(c)(3) non-profit organization in South Carolina, USA. == Current and previous conferences == 2024: Auckland, New Zealand (May 6-10) 2023: London, United Kingdom (May 29-June 1) 2022: Auckland, New Zealand (May 9–13) 2021: London, United Kingdom (May 3-May 7) 2020: Auckland, New Zealand (May 9–13) 2019: Montreal, Canada (May 13–17) 2018: Stockholm, Sweden (July 10–15) 2017: São Paulo, Brazil 2016: Singapore City, Singapore 2015: Istanbul, Turkey 2014: Paris, France 2013: Saint Paul, USA 2012: Valencia, Spain 2011: Taipei, Taiwan 2010: Toronto, Canada 2009: Budapest, Hungary 2008: Estoril, Portugal 2007: Honolulu, USA 2006: Hakodate, Japan 2005: Utrecht, The Netherlands 2004: New York, USA 2003: Melbourne, Australia 2002: Bologna, Italy == Activities == Besides the main program that consists of a main track, an industry and applications track, and a couple of special area tracks, AAMAS also hosts over 20 workshops (e.g., AOSE, COIN, DALT, ProMAS, to mention a few) and many tutorials. There is also a demonstration session and a doctoral symposium. Finally, each year AAMAS features a bunch of awards, most notably the IFAAMAS Influential Paper Award. It publishes proceedings which are available online.
The Raimones
The Raimones (stylized as THE RAiMONES) is a 2017 generative music project that utilized artificial intelligence to compose music in the style of the American punk rock band The Ramones. Developed by Matthias Frey, a researcher at Sony CSL Tokyo, the project was an early experiment in applying deep learning to high-energy, minimalist musical genres. == Technical Development == The project utilized Long short-term memory (LSTM) recurrent neural networks to generate musical structures and lyrics. The model was trained on a dataset consisting of 130 Ramones songs in MIDI format and the band's complete lyrical catalog. The technical framework was built using Python and Jupyter Notebook, drawing influence from the character-level RNN text generation models popularized by Andrej Karpathy. Unlike contemporary AI music projects that focused on the harmonic complexities of classical or pop music, THE RAiMONES sought to determine if neural networks could replicate the "1-2-3-4" rhythmic consistency and formulaic nature of early punk. == "I'm Alive" == The primary output of the project was the song "I'm Alive," released in 2017. The work is described as a form of "augmented intelligence," a hybrid approach where the AI provides the compositional foundation and human musicians handle the arrangement and performance. The song was recorded by the musician Mr. Ratboy (Gilbert Avondet). Avondet's involvement provided a stylistic link to the subject material, as he had previously served as a touring guitarist for Marky Ramone and the Intruders in 1996. The project's discography has since been made available on major streaming platforms, including Apple Music. == Reception and Significance == The project has been cited as a "proof of concept" for AI's ability to tackle "noisy" and aggressive aesthetics. In 2019, the Belgian magazine Knack Focus profiled the project alongside other AI pioneers such as Holly Herndon, noting the project's attempt to recreate the sound of "deceased legends" while maintaining a distinct, machine-like quality. It has also been featured in academic settings, such as at UC Santa Cruz, as a case study for AI-driven genre mimicry.
Interim Measures for the Management of Generative AI Services
The Interim Measures for the Management of Generative AI Services (Chinese: 生成式人工智能服务管理暂行办法; pinyin: Shēngchéng shì réngōng zhìnéng fúwù guǎnlǐ zànxíng bànfǎ) are a set of regulations governing public-facing generative artificial intelligence services in China. Issued on 10 July 2023 and effective from 15 August 2023, they were China's first binding regulation specifically targeting generative AI. They have been described as among the earliest such regulations adopted by any country. The measures were jointly issued by the Cyberspace Administration of China (CAC) and six other national bodies: the National Development and Reform Commission, the Ministry of Education, the Ministry of Science and Technology, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the National Radio and Television Administration. Among the measures' most prominent requirements is that generative AI services must uphold Core Socialist Values and must not generate content that could subvert state power, harm national security, or undermine social stability. The measures also require providers of public-facing generative AI services to undergo security assessments and register their algorithms with the CAC. As of December 2025, 748 generative AI services had completed the filing process at the national level. == Background == The Interim Measures build on two earlier sets of regulations targeting specific algorithm applications. The Administrative Provisions on Algorithm Recommendation for Internet Information Services, effective from March 2022, established China's algorithm registry and required providers of recommendation algorithms with "public opinion properties or social mobilization capabilities" to file with the CAC and undergo security assessments. The Administrative Provisions on Deep Synthesis of Internet Information Services, effective from January 2023, extended similar requirements to algorithms used for generating synthetic media such as deepfakes. In April 2023, the CAC released a draft of the generative AI regulation for public comment. The draft included several requirements that attracted attention, including that generated content should "embody Core Socialist Values" and that training data should be "true and accurate". The public consultation period ran until May 2023. The final version, published in July 2023, was substantially revised from the draft. According to an analysis by the Future of Privacy Forum, changes appeared to reflect feedback from industry stakeholders including Baidu, Xiaomi, SenseTime, and others, as well as input from government-affiliated research institutes. The final measures adopted a more permissive tone, with the CAC describing its approach as "inclusive and prudent" (包容审慎) and emphasising "classified and graded" (分类分级) supervision. == Scope == The measures apply to services that use generative AI technology to provide text, images, audio, video, or other content to the public within mainland China (Article 2). They do not apply to organisations that develop or use generative AI internally without offering services to the domestic public, such as industry associations, enterprises, and research institutions. Overseas providers whose services are accessible to users in China are also subject to the measures. == Key provisions == === Content requirements === Article 4 sets out the core content obligations. Providers and users of generative AI services must uphold the Core Socialist Values. The measures prohibit generating content that incites subversion of national sovereignty or the socialist system, endangers national security or the nation's image, incites separatism, promotes terrorism or extremism, promotes ethnic hatred or discrimination, or contains violence, obscenity, or false information prohibited by law. These content prohibitions largely mirror those in Article 12 of the Cybersecurity Law and in prior regulations governing online content. Article 4 also requires that models be designed and trained to avoid discrimination, that services respect intellectual property rights, and that providers take effective measures to improve the transparency and accuracy of generated content. === Training data and labelling === Article 7 requires providers to ensure that training data is of high quality and legitimately sourced, and that it does not infringe upon intellectual property rights. Where personal information is used, consent must be obtained. The final version of this provision removed language from the draft that would have held providers responsible for the "legitimacy" of all pretraining data, replacing it with a requirement to "employ effective measures to improve the quality of training data". Article 8 requires providers to establish labelling rules for training data and to conduct quality assessments of data annotations. Article 12 requires that generated images, videos, and other synthetic content be labelled as AI-generated. === User rights and privacy === Article 11 requires providers to protect user privacy, to minimise the collection and retention of personal data, and to refrain from unlawfully sharing user information. Users have the right to request review, correction, or deletion of their personal information. Article 10 requires providers to take measures to prevent excessive dependence on or addiction to generative AI services by minors. === Security assessment and algorithm filing === Article 17 requires that providers of generative AI services with "public opinion properties or the capacity for social mobilization" (具有舆论属性或者社会动员能力) carry out security assessments and complete algorithm filing procedures in accordance with the Administrative Provisions on Algorithm Recommendation for Internet Information Services. == Implementation == === Algorithm filing process === In practice, the filing requirements under the Interim Measures have developed into a two-tier process. The first tier is the standard algorithm filing (算法备案) under the pre-existing Algorithm Recommendation Provisions, which involves submitting information about an algorithm's design, purpose, and data sources to the CAC. This process is primarily a registration mechanism. For public-facing generative AI products, there is an additional, more rigorous process commonly referred to as the "large model filing" (大模型备案). This involves submitting a security self-assessment report, data annotation rules, a keyword blocking list, and evaluation test question sets. The process includes technical testing at the provincial level, followed by review at the national CAC level. The algorithm filing targets specific algorithms, while the large model filing evaluates the broader system architecture, training data, model parameters, and potential social impact. The CAC publishes lists of generative AI services that have successfully completed the filing process. The first such list was published on 2 April 2024. According to the CAC's year-end announcements, 302 generative AI services had completed national-level filing by the end of 2024 (of which 238 were new that year), alongside 105 applications that completed local-level registration. By the end of 2025, the cumulative total had risen to 748 national-level filings and 435 local-level registrations. === Content compliance and testing === According to the Carnegie Endowment, the CAC has conducted compliance audits of generative AI services with a particular focus on ensuring appropriate responses to queries about politically sensitive topics. The large model filing process requires providers to pass both provincial-level and national-level technical testing before their services can be made available to the public. On 1 March 2024, the National Technical Committee 260 on Cybersecurity (TC260) published TC260-003, the Basic Security Requirements for Generative AI Services (生成式人工智能服务安全基本要求), a technical standard that provides detailed guidance on the security assessments required under the Interim Measures. The standard covers requirements for training data safety, model security, and content safety evaluation, and is used as a reference for the filing process. == Analysis == === Relationship to broader Chinese internet regulation === The content requirements in the Interim Measures extend China's existing framework for online information control to generative AI. Legal scholars have noted that the "Core Socialist Values" provision and the specific content prohibitions are consistent with longstanding requirements imposed on internet platforms under the Cybersecurity Law and related regulations. The Asia Society Policy Institute has described the Chinese government's highest regulatory priority in this area as retaining control of information, noting that content-related obligations receive stricter enforcement than other provisions. === Nature of the filing system === The character of the filing system has been debated by scholars. Angela Huyue Zh
Gradient vector flow
Gradient vector flow (GVF), a computer vision framework introduced by Chenyang Xu and Jerry L. Prince, is the vector field that is produced by a process that smooths and diffuses an input vector field. It is usually used to create a vector field from images that points to object edges from a distance. It is widely used in image analysis and computer vision applications for object tracking, shape recognition, segmentation, and edge detection. In particular, it is commonly used in conjunction with active contour model. == Background == Finding objects or homogeneous regions in images is a process known as image segmentation. In many applications, the locations of object edges can be estimated using local operators that yield a new image called an edge map. The edge map can then be used to guide a deformable model, sometimes called an active contour or a snake, so that it passes through the edge map in a smooth way, therefore defining the object itself. A common way to encourage a deformable model to move toward the edge map is to take the spatial gradient of the edge map, yielding a vector field. Since the edge map has its highest intensities directly on the edge and drops to zero away from the edge, these gradient vectors provide directions for the active contour to move. When the gradient vectors are zero, the active contour will not move, and this is the correct behavior when the contour rests on the peak of the edge map itself. However, because the edge itself is defined by local operators, these gradient vectors will also be zero far away from the edge and therefore the active contour will not move toward the edge when initialized far away from the edge. Gradient vector flow (GVF) is the process that spatially extends the edge map gradient vectors, yielding a new vector field that contains information about the location of object edges throughout the entire image domain. GVF is defined as a diffusion process operating on the components of the input vector field. It is designed to balance the fidelity of the original vector field, so it is not changed too much, with a regularization that is intended to produce a smooth field on its output. Although GVF was designed originally for the purpose of segmenting objects using active contours attracted to edges, it has been since adapted and used for many alternative purposes. Some newer purposes including defining a continuous medial axis representation, regularizing image anisotropic diffusion algorithms, finding the centers of ribbon-like objects, constructing graphs for optimal surface segmentations, creating a shape prior, and much more. == Theory == The theory of GVF was originally described by Xu and Prince. Let f ( x , y ) {\displaystyle \textstyle f(x,y)} be an edge map defined on the image domain. For uniformity of results, it is important to restrict the edge map intensities to lie between 0 and 1, and by convention f ( x , y ) {\displaystyle \textstyle f(x,y)} takes on larger values (close to 1) on the object edges. The gradient vector flow (GVF) field is given by the vector field v ( x , y ) = [ u ( x , y ) , v ( x , y ) ] {\displaystyle \textstyle \mathbf {v} (x,y)=[u(x,y),v(x,y)]} that minimizes the energy functional In this equation, subscripts denote partial derivatives and the gradient of the edge map is given by the vector field ∇ f = ( f x , f y ) {\displaystyle \textstyle \nabla f=(f_{x},f_{y})} . Figure 1 shows an edge map, the gradient of the (slightly blurred) edge map, and the GVF field generated by minimizing E {\displaystyle \textstyle {\mathcal {E}}} . Equation 1 is a variational formulation that has both a data term and a regularization term. The first term in the integrand is the data term. It encourages the solution v {\displaystyle \textstyle \mathbf {v} } to closely agree with the gradients of the edge map since that will make v − ∇ f {\displaystyle \textstyle \mathbf {v} -\nabla f} small. However, this only needs to happen when the edge map gradients are large since v − ∇ f {\displaystyle \textstyle \mathbf {v} -\nabla f} is multiplied by the square of the length of these gradients. The second term in the integrand is a regularization term. It encourages the spatial variations in the components of the solution to be small by penalizing the sum of all the partial derivatives of v {\displaystyle \textstyle \mathbf {v} } . As is customary in these types of variational formulations, there is a regularization parameter μ > 0 {\displaystyle \textstyle \mu >0} that must be specified by the user in order to trade off the influence of each of the two terms. If μ {\displaystyle \textstyle \mu } is large, for example, then the resulting field will be very smooth and may not agree as well with the underlying edge gradients. Theoretical Solution. Finding v ( x , y ) {\displaystyle \textstyle \mathbf {v} (x,y)} to minimize Equation 1 requires the use of calculus of variations since v ( x , y ) {\displaystyle \textstyle \mathbf {v} (x,y)} is a function, not a variable. Accordingly, the Euler equations, which provide the necessary conditions for v {\displaystyle \textstyle \mathbf {v} } to be a solution can be found by calculus of variations, yielding where ∇ 2 {\displaystyle \textstyle \nabla ^{2}} is the Laplacian operator. It is instructive to examine the form of the equations in (2). Each is a partial differential equation that the components u {\displaystyle u} and v {\displaystyle v} of v {\displaystyle \mathbf {v} } must satisfy. If the magnitude of the edge gradient is small, then the solution of each equation is guided entirely by Laplace's equation, for example ∇ 2 u = 0 {\displaystyle \textstyle \nabla ^{2}u=0} , which will produce a smooth scalar field entirely dependent on its boundary conditions. The boundary conditions are effectively provided by the locations in the image where the magnitude of the edge gradient is large, where the solution is driven to agree more with the edge gradients. Computational Solutions. There are two fundamental ways to compute GVF. First, the energy function E {\displaystyle {\mathcal {E}}} itself (1) can be directly discretized and minimized, for example, by gradient descent. Second, the partial differential equations in (2) can be discretized and solved iteratively. The original GVF paper used an iterative approach, while later papers introduced considerably faster implementations such as an octree-based method, a multi-grid method, and an augmented Lagrangian method. In addition, very fast GPU implementations have been developed in Extensions and Advances. GVF is easily extended to higher dimensions. The energy function is readily written in a vector form as which can be solved by gradient descent or by finding and solving its Euler equation. Figure 2 shows an illustration of a three-dimensional GVF field on the edge map of a simple object (see ). The data and regularization terms in the integrand of the GVF functional can also be modified. A modification described in , called generalized gradient vector flow (GGVF) defines two scalar functions and reformulates the energy as While the choices g ( ∇ f | ) = μ {\displaystyle \textstyle g(\nabla f|)=\mu } and h ( | ∇ f | ) = | ∇ f | 2 {\displaystyle \textstyle h(|\nabla f|)=|\nabla f|^{2}} reduce GGVF to GVF, the alternative choices g ( | ∇ f | ) = exp { − | ∇ f | / K } {\displaystyle \textstyle g(|\nabla f|)=\exp\{-|\nabla f|/K\}} and h ( ∇ f | ) = 1 − g ( | ∇ f | ) {\displaystyle \textstyle h(\nabla f|)=1-g(|\nabla f|)} , for K {\displaystyle K} a user-selected constant, can improve the tradeoff between the data term and its regularization in some applications. The GVF formulation has been further extended to vector-valued images in where a weighted structure tensor of a vector-valued image is used. A learning based probabilistic weighted GVF extension was proposed in to further improve the segmentation for images with severely cluttered textures or high levels of noise. The variational formulation of GVF has also been modified in motion GVF (MGVF) to incorporate object motion in an image sequence. Whereas the diffusion of GVF vectors from a conventional edge map acts in an isotropic manner, the formulation of MGVF incorporates the expected object motion between image frames. An alternative to GVF called vector field convolution (VFC) provides many of the advantages of GVF, has superior noise robustness, and can be computed very fast. The VFC field v V F C {\displaystyle \textstyle \mathbf {v} _{\mathrm {VFC} }} is defined as the convolution of the edge map f {\displaystyle f} with a vector field kernel k {\displaystyle \mathbf {k} } where The vector field kernel k {\displaystyle \textstyle \mathbf {k} } has vectors that always point toward the origin but their magnitudes, determined in detail by the function m {\displaystyle m} , decrease to zero with increasing distance from the origin. The beauty of VFC is that it can be computed very rapidly using a fast Fourier tra
Stable Diffusion
Stable Diffusion is a deep learning, text-to-image model released in 2022 based on diffusion techniques. The generative artificial intelligence technology is the premier product of Stability AI and is considered to be a part of the ongoing AI boom. It is primarily used to generate detailed images conditioned on text descriptions, though it can also be applied to other tasks such as inpainting, outpainting, and generating image-to-image translations guided by a text prompt. Its development involved researchers from the CompVis Group at LMU Munich and Runway with a computational donation from Stability and training data from non-profit organizations. Stable Diffusion is a latent diffusion model, a kind of deep generative artificial neural network. Its code and model weights have been released publicly, and an optimized version can run on most consumer hardware equipped with a modest GPU with as little as 2.4 GB VRAM. This marked a departure from previous proprietary text-to-image models such as DALL-E and Midjourney which were accessible only via cloud services. == Development == Stable Diffusion originated from a project called Latent Diffusion, developed in Germany by researchers at LMU Munich in Munich and Heidelberg University. Four of the original 5 authors (Robin Rombach, Andreas Blattmann, Patrick Esser and Dominik Lorenz) later joined Stability AI and released subsequent versions of Stable Diffusion. The technical license for the model was released by the CompVis group at LMU Munich. Development was led by Patrick Esser of Runway and Robin Rombach of CompVis, who were among the researchers who had earlier invented the latent diffusion model architecture used by Stable Diffusion. Stability AI also credited EleutherAI and LAION (a German nonprofit which assembled the dataset on which Stable Diffusion was trained) as supporters of the project. == Technology == === Architecture === Diffusion models, introduced in 2015, are trained with the objective of removing successive applications of Gaussian noise on training images, which can be thought of as a sequence of denoising autoencoders. The name diffusion is from the thermodynamic diffusion, since they were first developed with inspiration from thermodynamics. Models in Stable Diffusion series before SD 3 all used a variant of diffusion models, called latent diffusion model (LDM), developed in 2021 by the CompVis (Computer Vision & Learning) group at LMU Munich. Stable Diffusion consists of 3 parts: the variational autoencoder (VAE), U-Net, and an optional text encoder. The VAE encoder compresses the image from pixel space to a smaller dimensional latent space, capturing a more fundamental semantic meaning of the image. Gaussian noise is iteratively applied to the compressed latent representation during forward diffusion. The U-Net block, composed of a ResNet backbone, denoises the output from forward diffusion backwards to obtain a latent representation. Finally, the VAE decoder generates the final image by converting the representation back into pixel space. The denoising step can be flexibly conditioned on a string of text, an image, or another modality. The encoded conditioning data is exposed to denoising U-Nets via a cross-attention mechanism. For conditioning on text, the fixed, pretrained CLIP ViT-L/14 text encoder is used to transform text prompts to an embedding space. Researchers point to increased computational efficiency for training and generation as an advantage of LDMs. With 860 million parameters in the U-Net and 123 million in the text encoder, Stable Diffusion is considered relatively lightweight by 2022 standards, and unlike other diffusion models, it can run on consumer GPUs, and even CPU-only if using the OpenVINO version of Stable Diffusion. ==== SD XL ==== The XL version uses the same LDM architecture as previous versions, except larger: larger UNet backbone, larger cross-attention context, two text encoders instead of one, and trained on multiple aspect ratios (not just the square aspect ratio like previous versions). The SD XL Refiner, released at the same time, has the same architecture as SD XL, but it was trained for adding fine details to preexisting images via text-conditional img2img. ==== SD 3.0 ==== The 3.0 version completely changes the backbone. Not a UNet, but a Rectified Flow Transformer, which implements the rectified flow method with a Transformer. The Transformer architecture used for SD 3.0 has three "tracks", for original text encoding, transformed text encoding, and image encoding (in latent space). The transformed text encoding and image encoding are mixed during each transformer block. The architecture is named "multimodal diffusion transformer (MMDiT), where the "multimodal" means that it mixes text and image encodings inside its operations. This differs from previous versions of DiT, where the text encoding affects the image encoding, but not vice versa. === Training data === Stable Diffusion was trained on pairs of images and captions taken from LAION-5B, a publicly available dataset derived from Common Crawl data scraped from the web, where 5 billion image-text pairs were classified based on language and filtered into separate datasets by resolution, a predicted likelihood of containing a watermark, and predicted "aesthetic" score (e.g. subjective visual quality). The dataset was created by LAION, a German non-profit which receives funding from Stability AI. The Stable Diffusion model was trained on three subsets of LAION-5B: laion2B-en, laion-high-resolution, and laion-aesthetics v2 5+. A third-party analysis of the model's training data identified that out of a smaller subset of 12 million images taken from the original wider dataset used, approximately 47% of the sample size of images came from 100 different domains, with Pinterest taking up 8.5% of the subset, followed by websites such as WordPress, Blogspot, Flickr, DeviantArt and Wikimedia Commons. An investigation by Bayerischer Rundfunk showed that LAION's datasets, hosted on Hugging Face, contain large amounts of private and sensitive data. === Training procedures === The model was initially trained on the laion2B-en and laion-high-resolution subsets, with the last few rounds of training done on LAION-Aesthetics v2 5+, a subset of 600 million captioned images which the LAION-Aesthetics Predictor V2 predicted that humans would, on average, give a score of at least 5 out of 10 when asked to rate how much they liked them. The LAION-Aesthetics v2 5+ subset also excluded low-resolution images and images which LAION-5B-WatermarkDetection identified as carrying a watermark with greater than 80% probability. Final rounds of training additionally dropped 10% of text conditioning to improve Classifier-Free Diffusion Guidance. The model was trained using 256 Nvidia A100 GPUs on Amazon Web Services for a total of 150,000 GPU-hours, at a cost of $600,000. === Limitations === Stable Diffusion has issues with degradation and inaccuracies in certain scenarios. Initial releases of the model were trained on a dataset that consists of 512×512 resolution images, meaning that the quality of generated images noticeably degrades when user specifications deviate from its "expected" 512×512 resolution; the version 2.0 update of the Stable Diffusion model later introduced the ability to natively generate images at 768×768 resolution. Another challenge is in generating human limbs due to poor data quality of limbs in the LAION database. The model is insufficiently trained to replicate human limbs and faces due to the lack of representative features in the database, and prompting the model to generate images of such type can confound the model. In addition to human limbs, Stable Diffusion is unable to generate legible ambigrams and some other forms of text and typography. Stable Diffusion XL (SDXL) version 1.0, released in July 2023, introduced native 1024x1024 resolution and improved generation for limbs and text. Accessibility for individual developers can also be a problem. In order to customize the model for new use cases that are not included in the dataset, such as generating anime characters ("waifu diffusion"), new data and further training are required. Fine-tuned adaptations of Stable Diffusion created through additional retraining have been used for a variety of different use-cases, from medical imaging to algorithmically generated music. However, this fine-tuning process is sensitive to the quality of new data; low resolution images or different resolutions from the original data can not only fail to learn the new task but degrade the overall performance of the model. Even when the model is additionally trained on high quality images, it is difficult for individuals to run models in consumer electronics. For example, the training process for waifu-diffusion requires a minimum 30 GB of VRAM, which exceeds the usual resource provided in such consumer GPUs as Nvidia's GeForce 30 series, w