About
I am a research leader in Machine Learning and a Computer Vision researcher.
I…
I…
Services
Activity
11K followers
-
Anton van den Hengel reposted thisAnton van den Hengel reposted thisOn 30 March, Kingston AI Group Chair and AIML Chief Scientist Professor Anton van den Hengel joined an expert panel at The Deep Tech Founder Salon in Sydney, Anthropic's first-ever Australian event, hosted by Cicada Innovations in partnership with Main Sequence Ventures. The event brought together deep tech founders, venture capitalists, researchers, and members of the academic community and press to explore the role AI is already playing in frontier research and industry. The panel and fireside chat, composed of Anthropic Technical Staff Zac Hatfield-Dodds; In The Blink of AI Podcast Host Georgie Healy; Neara CEO Daniel Danilatos; Harrison.ai Engineering Head Dr Suneeta Mall; Presien CEO Mark Richards; and Professor van den Hengel examined how AI is being applied inside frontier companies today, with a particular focus on safety and trust in regulated domains. The conversation, moderated by Main Sequence Ventures Investment Manager Danielle Haj-Moussa, challenged the notion that safety and regulation are barriers and instead reframed them as competitive advantages for those building in high-stakes industries like healthcare, infrastructure, and climate. Professor van den Hengel drew on his research at AIML to speak to the real-world impact of AI and the opportunities it presents for Australian deep tech. "It was fantastic to see the depth of Australian deep tech AI talent at the event," he said. We congratulate Cicada Innovations, Main Sequence Ventures, and Anthropic on a landmark event, and looks forward to continued collaboration with Australia's deep tech community. Photo captions: (L-R) Georgie Healy, Zac Hatfield-Dodds, Mark Richards, Daniel Danilatos, Dr Suneeta Mall, Professor Anton van den Hengel, and Danielle Haj-Moussa at The Deep Tech Founder Salon event. | (Left) Danielle Haj-Moussa hosting the panel conversation. | (Right) Professor Anton van den Hengel addressing the audience. Photos provided by Remarkable Communications.
-
Anton van den Hengel shared thisMuch of the narrative around AI ignores the fact that we have the option to use it for purposes other than making money for multinationals. One application might be to build metric measures of biodiversity to drive investment in protecting the environment.Anton van den Hengel shared thisIt was a great privilege to be part of a team including Adrian Ward from Accounting for Nature and Anton van den Hengel from Australian Institute for Machine Learning (AIML) to present to 50 leaders from the Asian Development Bank (ADB) on a proposal to develop low-cost, high-trust measurement tools for habitat quality using #AI. The intention is to create greater trust in nature finance markets to help address the funding gap for habitat protection and restoration in the Asian Pacific Region. The Rozetta Institute are enormously grateful to Emil Bolongaita and his team for hosting us and to Adrian and Anton for developing and co-presenting the proposal. #nature #finance #habitat #restoration #protection #investment Daisy Mallett Hugh Possingham Tim Moore PhD Prof Ariella Helfgott Peter Cosier AM
-
Anton van den Hengel reposted thisAnton van den Hengel reposted thisTwo years ago, the Productivity Commission told us that AI was essential to Australia's productivity but that we should just buy it off the shelf. 📦💻 Yesterday, Assistant Minister Dr Andrew Charlton reframed the debate: “The question is not whether AI lifts productivity. It will. The question is who captures the gains.” For years many of us, like the Kingston AI Group, have argued that Australia must back its own AI capability. We have the talent. We have the research depth. We have the energy advantage. What we’ve lacked is urgency and ambition. Now there’s finally a shift in Australia from adoption to creation, which I write about in this article. 🚀 #AI #Productivity #AustralianAI #SovereignAI #Innovation #AustralianTech #DigitalEconomy #madeinAustralia
-
Anton van den Hengel reposted thisAnton van den Hengel reposted thisFor most of the 2010s, convolutional neural networks (ConvNets) — networks that process images by sliding small filters across them to detect patterns — were the dominant architecture for computer vision. Then transformers arrived. Transformers, originally designed for text, process input by letting every element attend to every other element simultaneously, and when adapted for images they started outperforming ConvNets on benchmark after benchmark. The general consensus became that transformers were just better, and ConvNets were the old way. This paper's authors decided to test whether that conclusion was actually justified, or whether ConvNets had simply fallen behind in terms of training recipes and design choices rather than fundamental capability. They took a standard ResNet and systematically updated it — one change at a time — borrowing design decisions from transformers: things like which activation function to use, which type of normalization to apply, and how to reshape the internal dimensions of each processing block. Each change was small and interpretable, and the paper tracks the accuracy at every step so you can see exactly what each decision contributes. The final result is a pure ConvNet that matches or beats transformer-based models of equivalent size on image classification, object detection, and segmentation, by being up to 49% faster. Read with AI tutor: https://lnkd.in/eXRDBcW5 Read alone: https://lnkd.in/euBG698R
-
Anton van den Hengel reposted thisLet me know if using this tool saves you money 💸 Comment yes or no, or just start a brawl with bystanders (standard internet courtesy) Dont guess download the data to be sure. https://lnkd.in/e258UwzEAnton van den Hengel reposted thisSo lets get some participation! Would it cost less to use this data or start from nothing? Download Nuke 16 file and Import Script and Raw Data 💾 https://lnkd.in/evAtDJCr
-
Anton van den Hengel reposted thisAnton van den Hengel reposted this🔥 𝗪𝗵𝗮𝘁 𝗶𝗳 𝘄𝗲𝗯 𝘁𝗲𝘅𝘁 𝗶𝘀𝗻’𝘁 𝘁𝗵𝗲 𝗯𝗲𝘀𝘁 𝗽𝗹𝗮𝗰𝗲 𝘁𝗼 𝘀𝘁𝗮𝗿𝘁 𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 𝗟𝗟𝗠𝘀? In our latest paper, we explore a simple idea: 👉 Before exposing models to massive amounts of semantic web data, what if we first warm them up on structured *abstract* data? Across models up to 1.3B parameters, we find that pretraining on procedural data (formal languages, simple algorithms, etc.) accelerates and improves subsequent training on language, code, and math. 💡 𝗪𝗵𝘆 𝗽𝗿𝗼𝗰𝗲𝗱𝘂𝗿𝗮𝗹 𝗱𝗮𝘁𝗮? Humans often learn better by first mastering simple structure and logic, rather than memorizing vast amounts of facts. By analogy, we use abstract, structured data to build a scaffold, free of semantic biases. 🔍 𝗗𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗽𝗿𝗼𝗰𝗲𝗱𝘂𝗿𝗮𝗹 𝗱𝗮𝘁𝗮 → 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗯𝗲𝗻𝗲𝗳𝗶𝘁𝘀 Using algorithmic diagnostic tasks, we observe distinct effects, e.g.: • Some procedural data improves long-context recall, • Others boost arithmetic ability. 📈 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀 𝗱𝘂𝗿𝗶𝗻𝗴 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱 𝗽𝗿𝗲𝘁𝗿𝗮𝗶𝗻𝗶𝗻𝗴 Just 0.1% of procedural data, front-loaded before standard pretraining, leads to: • Significantly better performance on language, code, and math, • Up to 45% less semantic data required to reach a baseline perplexity. ⚙️ 𝗠𝗟𝗣𝘀 𝘃𝘀. 𝗔𝘁𝘁𝗲𝗻𝘁𝗶𝗼𝗻: 𝘄𝗵𝗲𝗿𝗲 𝗱𝗼 𝘁𝗵𝗲 𝗯𝗲𝗻𝗲𝗳𝗶𝘁𝘀 𝗹𝗶𝘃𝗲? We ran weight-reset experiments before standard pretraining. Surprisingly, resetting selected weights can bring further gains, but in a domain-specific way: • Warmed-up MLPs → better natural language • Warmed-up attention → better code & math This suggests that different architectural components internalize different kinds of procedural structure. 🧩 𝗠𝗶𝘅𝗶𝗻𝗴 𝗵𝗲𝗹𝗽𝘀 We get further gains by either: • Combining multiple types of procedural data, or • Mixing weights from models individually warmed up on different types of data. ──────────────────────────────────────────────────────────── There’s a lot more to explore here around inductive biases and data curriculum design. Curious to hear thoughts from others working on pretraining efficiency, scaling, and architecture optimization! This is joint work with Liangze Jiang, Zachary Shinnick, Anton van den Hengel, Hemanth Saratchandran. 📄 Paper on arXiv: https://lnkd.in/eecs78WJ 💻 Code on GitHub: https://lnkd.in/eZiu37wC
-
Anton van den Hengel reposted thisAnton van den Hengel reposted this🚀 What an incredible milestone! Delighted to congratulate the SHARON AI team on debuting on the Nasdaq with a ~A$1 billion valuation (NAS: SHAZ) — a huge testament to the vision, hard work and execution of everyone involved. Thank you to our talented team, dedicated partners, supportive investors, and early adopters — this achievement wouldn’t be possible without you. Onwards and upwards! Looking forward to what’s next as we continue to innovate and scale globally. $SHAZ https://lnkd.in/gDV52jE8Firmus rival Sharon AI debuts with $1b valuation on NasdaqFirmus rival Sharon AI debuts with $1b valuation on Nasdaq
-
Anton van den Hengel reposted thisAnton van den Hengel reposted thisDear Mr. Treasurer, earlier this week, we presented to your colleagues at Reserve Bank of Australia, Australian Prudential Regulation Authority, Treasury and ASIC on the Economic risks of AI. "You cannot divest from the algorithm". When I did a global AI LLM ecosystem map and explained that the Saudi Arabian Federal Government, Public Infrastructure Fund, had invested USD$100Bn in their LLM and AI infrastructure and that their National Security Chief was Chair of the Board of their LLM, a few eyebrows were raised. Think about this for a minute. "Middle powers must act together because if we’re not at the table, we’re on the menu." - Mark Carney, WEF2026, Davos. Only one ASIC leader understood the geopolitical risk and fiscal policy severity prior to our presentation. Not surprising, he was an American who studied at MIT. He asked the only pertinent question, "How do we regulate against the AI and what is best governance?" I said, "Fantastic question Joe, we CAN'T, unless we build our own (LLMs)". Joe gets it, thank you Joe. The rest of the room including the Head of Digital Assets of a bank were digesting it. I think I'm moving my retirement savings into a new fund unless Australian Superannuation and the policymakers act. This presentation was 6 months of work by Gordon Noble and myself in the interests of the Australian Economy, and the safety of Australian citizens. Where to from here? Susanne (Sue) Dahn AM Susan Oliver AM Janette Kendall AM Ming Long AM Bridget Loudon-Harris Nicola Roxon Alison Kitchen AM Nisha Padmanabhan Dr Katherine Woodthorpe AO FTSE FAICD Sarah Pearson Andrea Durrant Jo Mikleus Vanessa (Fernandes) Sullivan Monique Conheady Monique Bachner Elanor Huntington Do the female directors in the nation want to be the next Chairman of the Board? #SamMostyn pls share with your peers. Unlike Climate Risk Disclosure that took 3 decades, we don't have the luxury this time around (nor did we then). We need a Digital Governance Reporting Framework as a first step for 'active stewardship'. The Directors of this nation do not understand these risks. We all need to up our game. If you want to learn more, we will be hosting a Global AI Governance for Directors session in March.
-
Anton van den Hengel reposted thisAnton van den Hengel reposted thisOn this International Day of Women and Girls in Science, we’re celebrating the women shaping the future of machine learning research at AIML. In this Researcher Spotlight, Madeleine Cochrane, PhD student at AIML and the Institute for Photonics and Advanced Sensing (IPAS) at Adelaide University, shares her work on developing a breathalyser designed to detect disease. Her research sits at the intersection of photonics and machine learning. By shining lasers through breath samples, researchers capture complex light signals that act like chemical fingerprints. Madeleine develops machine learning models that decode those signals, helping identify which chemicals are present in the breath and in what concentration. “The human breath is actually a fantastic resource for getting an insight into what's going on inside your body,” said Madeleine. “Every exhaled breath contains thousands of different chemicals, and scientists have been able to link several of those chemicals to a range of diseases.” Because breath testing is completely non-invasive, the technology could support earlier screening for conditions such as cancer and to provide safer alternatives for patients who cannot tolerate invasive tests, highlighting the kind of real-world benefit that motivates Madeleine’s research. “What inspires me is the thought of the potential good that it can bring to people,” she said. “There is so much scope for benefits to so many different people.” 🎞️ Watch the full interview below (captions available) or on YouTube here: https://lnkd.in/gHxuMf8b #Photonics #MachineLearning #MedicalDiagnostics #InternationalDayOfWomenAndGirlsInScience Johan Verjans MD PhD FESC FRACP Zhibin LiaoAI for detecting disease from a single breath | Researcher spotlight – Madeleine CochraneAI for detecting disease from a single breath | Researcher spotlight – Madeleine Cochrane
Anton van den Hengel commented on a post
3w
The Saspockalipse is a great opportunity for us, but we need to lean in. The narrative from Atlassian has been that they sacked 1600 people due to AI. The real reason they sacked 1600 people is a lack of AI. Their share price plummeted because they didn't have a convincing plan for how to use AI to grow their business, and they had to sack people to reduce costs. The fact that they are saying they will replace them with AI is irrelevant. The Saspockalipse is affecting a host of companies that don't have a real AI plan.
Anton van den Hengel commented on a post
1mo
Great paper! What happens in the limit? Can we just do a series of rank 1 updates?
Anton van den Hengel replied to a comment
1mo
Sarah McRae I've got my fingers crossed too, but we've got a long way to go yet.
-
Anton van den Hengel liked thisAnton van den Hengel liked thisYesterday I spoke at the launch of the UK Housing Review 2026 at the House of Lords in London. In this year’s Review, Amy Clair and I contributed a chapter on housing and health. In it, our argument is simple: housing is not just a market issue or a policy issue — it is a public health issue. Affordability shapes whether people can avoid cold, damp and overcrowded homes. Poor housing conditions do not just make life harder — they damage health. One of the strongest messages from the evidence is that tenure insecurity is health harming. Worrying about eviction, rent, bills or repairs is not a minor background pressure - it leaves a real mark. We suggest that housing policy should be judged not only by supply numbers or market outcomes, but by whether it delivers healthy, secure, affordable and suitable homes. Thank you to the editors and organisers of the UK Housing Review for bringing together such an important publication and event. Download your copy here… https://lnkd.in/gTC9CQY9 #UKHR26
-
Anton van den Hengel liked thisAnton van den Hengel liked thisOn 30 March, Kingston AI Group Chair and AIML Chief Scientist Professor Anton van den Hengel joined an expert panel at The Deep Tech Founder Salon in Sydney, Anthropic's first-ever Australian event, hosted by Cicada Innovations in partnership with Main Sequence Ventures. The event brought together deep tech founders, venture capitalists, researchers, and members of the academic community and press to explore the role AI is already playing in frontier research and industry. The panel and fireside chat, composed of Anthropic Technical Staff Zac Hatfield-Dodds; In The Blink of AI Podcast Host Georgie Healy; Neara CEO Daniel Danilatos; Harrison.ai Engineering Head Dr Suneeta Mall; Presien CEO Mark Richards; and Professor van den Hengel examined how AI is being applied inside frontier companies today, with a particular focus on safety and trust in regulated domains. The conversation, moderated by Main Sequence Ventures Investment Manager Danielle Haj-Moussa, challenged the notion that safety and regulation are barriers and instead reframed them as competitive advantages for those building in high-stakes industries like healthcare, infrastructure, and climate. Professor van den Hengel drew on his research at AIML to speak to the real-world impact of AI and the opportunities it presents for Australian deep tech. "It was fantastic to see the depth of Australian deep tech AI talent at the event," he said. We congratulate Cicada Innovations, Main Sequence Ventures, and Anthropic on a landmark event, and looks forward to continued collaboration with Australia's deep tech community. Photo captions: (L-R) Georgie Healy, Zac Hatfield-Dodds, Mark Richards, Daniel Danilatos, Dr Suneeta Mall, Professor Anton van den Hengel, and Danielle Haj-Moussa at The Deep Tech Founder Salon event. | (Left) Danielle Haj-Moussa hosting the panel conversation. | (Right) Professor Anton van den Hengel addressing the audience. Photos provided by Remarkable Communications.
-
Anton van den Hengel liked thisAnton van den Hengel liked thisAIML at Adelaide University is hiring three talented and driven AI researchers to join the institute. 🟣 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐅𝐞𝐥𝐥𝐨𝐰 (𝐁) – 𝐈𝐧𝐝𝐮𝐬𝐭𝐫𝐢𝐚𝐥 𝐀𝐈 𝐏𝐫𝐨𝐠𝐫𝐚𝐦 (2x 𝐩𝐨𝐬𝐢𝐭𝐢𝐨𝐧𝐬) Funded by the SA Government to drive economic growth and job creation, the Industrial AI Program is a major state-supported initiative advancing core capability in industrial AI across South Australia and nationally. Working with Associate Professor Xin Yu, successful candidates will: ✔️ Have a PhD in computer science, engineering, or a related discipline; ✔️ Conduct high-impact research in industrial AI, contributing to leading publications; ✔️ Develop and deploy AI solutions to real-world challenges across key sectors; ✔️ Collaborate with leading researchers, engineers, and industry partners; and ✔️ Be part of a dynamic, multidisciplinary team within one of Australia's largest machine learning research institutes. Applications close 20 April 2026. For more information and to apply, visit: https://lnkd.in/gkbXtydD 🟡 𝐏𝐨𝐬𝐭𝐝𝐨𝐜𝐭𝐨𝐫𝐚𝐥 𝐑𝐞𝐬𝐞𝐚𝐫𝐜𝐡𝐞𝐫 (𝐀) – 𝐍𝐞𝐱𝐭-𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐀𝐈 & 𝐕𝐢𝐬𝐢𝐨𝐧-𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 Join a funded research project at the forefront of Large Vision-Language Models (LVLMs), developing innovative evaluation tools and training methods to enhance complex reasoning, decision-making, and adaptability in next-generation AI systems. Work that will address critical limitations in current AI and contribute to real-world solutions across healthcare, business, and education. Working with Associate Professor Qi Wu, successful candidates will have: ✔️ A PhD in machine learning, AI, computer science, or a related field; ✔️ Programming experience in Python, C++ or similar, with deep neural network frameworks; ✔️ Demonstrable expertise in deep learning, computer vision, LLMs, or related areas; and ✔️ A strong publication record in top-tier machine learning or computer vision venues. Applications close 23 April 2026. For more information and to apply, visit: https://lnkd.in/g7vwUjfd
-
Anton van den Hengel liked thisAnton van den Hengel liked thisIt was a great privilege to be part of a team including Adrian Ward from Accounting for Nature and Anton van den Hengel from Australian Institute for Machine Learning (AIML) to present to 50 leaders from the Asian Development Bank (ADB) on a proposal to develop low-cost, high-trust measurement tools for habitat quality using #AI. The intention is to create greater trust in nature finance markets to help address the funding gap for habitat protection and restoration in the Asian Pacific Region. The Rozetta Institute are enormously grateful to Emil Bolongaita and his team for hosting us and to Adrian and Anton for developing and co-presenting the proposal. #nature #finance #habitat #restoration #protection #investment Daisy Mallett Hugh Possingham Tim Moore PhD Prof Ariella Helfgott Peter Cosier AM
-
Anton van den Hengel liked thisAnton van den Hengel liked thisThe Kingston AI Group would like to extend our deepest thanks to all presenters and sponsors of our inaugural Kingston AI Group symposium, held Thursday, 19 March 2026 at the Adelaide Oval. Australian Institute for Machine Learning (AIML) Manager Dr Kathy Nicholson served as master of ceremonies for the event which featured keynote presentations from Kingston AI Group members, Adelaide University Professor Anton van den Hengel and Professor Dana Kulic of Monash University. Attendees were also treated to research presentations from Minh Hoang Nguyen and Hung Le from Deakin University; AIML's Townim Faisal Chowdhury, Yiping Ji, and Hemanth Saratchandran; Evan M. of the The Australian National University; Jiaheng Dong from the University of Melbourne; and Yue Yang from Maincode. Our thanks to sponsors Google Australia, ResetData, and Adelaide University for their support of this seminal new event. In our inaugural symposium in 2026, we were joined by approximately 150 AI researchers, practitioners, and enthusiasts, and we know this event will only grow in size and impact over the coming years. For more on the Kingston AI Symposium, please visit ➡️ https://lnkd.in/gSk6fc8q
-
Anton van den Hengel liked thisIt took me 4 years to get to _SOUTHSTART, which was 4 years too long! It was wonderful to hang out with the marvellous Bilue crew - Cameron Barrie and Ruth Neech. The event was like having a giant geodesic (representing the locally shortest path between two points on a surface) mirror held up to your face. It wasn't SXSW but it is Australia's answer to seeing the future, now. Big shoutout to the people I had the honour of deepening connections with... I still have to get back to all of you and plan our next collaboration/catchup. Dr Milad D. Anton van den Hengel Base Yates Craig Swann Eike Zeller ☀️ Christopher Sassone Whitney Cali Andrew Lewan Troy Smithells Rod Hamilton Martin Boyd Eloise LeaverAnton van den Hengel liked thisI spent last week with a collective of amazing humans doing innovative things. No SXSW for me this year, it was SOUTHSTART instead. Thanks to all those I connected with and leart from, you were all amazing. Here's my thoughts on the event, and why, in the words of the governator, I'll be back.Why I Ditched SXSW for Adelaide (And I'd Do It Again)Why I Ditched SXSW for Adelaide (And I'd Do It Again)
-
Anton van den Hengel liked thisAnton van den Hengel liked thisEarlier this week, we gathered at Cicada Innovations for Anthropic’s first-ever event in Sydney. 🇦🇺 Hosting Anthropic’s first-ever Aussie event was a landmark moment, but the real story is their deliberate alignment with deep tech startups, announcing up to $50k USD in Claude tokens specifically for the teams using AI as a foundational engine for scientific discovery. This commitment is a powerful indicator of Australia’s growing prominence in the global landscape; world-class AI safety research is moving directly into the path of those building our most essential systems. Thank you to the speakers and panellists: Zac Hatfield-Dodds (Anthropic), Georgie Healy (In The Blink of AI), Daniel Danilatos (Neara), Suneeta Mall (Harrison.ai), Mark Richards (Presien), and Anton van den Hengel (Australian Institute for Machine Learning (AIML)) for reminding us that when we build for reality, scale, and consequence, safety and regulation become competitive advantages. Thank you also to our partners, our brave founders, and the 200+ pioneers who joined us. The future is being built here. 🤝 #DeepTech #AI
Experience & Education
-
Australian Institute for Machine Learning (AIML)
View Anton’s full experience
See their title, tenure and more.
By clicking Continue to join or sign in, you agree to LinkedIn’s User Agreement, Privacy Policy, and Cookie Policy.
Publications
Patents
-
Method and software for analysing microbial growth
Issued US US9292729B2
A method for analysing microbial growth on a solid culture medium, the method including obtaining image data of the solid culture medium and any microbial growth, generating an associated feature vector of values obtained by applying one or more filters to the image data, using a classifier to classify each pixel in a plurality of pixels in the image data based on the associated feature vector, analysing results of pixel classifications of each said pixel to derive a microbiological assessment…
A method for analysing microbial growth on a solid culture medium, the method including obtaining image data of the solid culture medium and any microbial growth, generating an associated feature vector of values obtained by applying one or more filters to the image data, using a classifier to classify each pixel in a plurality of pixels in the image data based on the associated feature vector, analysing results of pixel classifications of each said pixel to derive a microbiological assessment of the solid culture medium and any microbial growth, and outputting the microbiological assessment.
Other inventors -
-
Method and system for generating a 3D model
Issued US US8675951B2
A method for generating a three dimensional (3D) model of an object is depicted in a two dimensional (2D) image is disclosed, The 2D image includes associated 3D information. The method includes an operator determining a geometric primitive corresponding to shape characteristics of the object and then the generation of the 3D model based on the determined geometric primitive, the 2D image, and the associated 3D information.
Other inventors -
-
Image processing method and apparatus
JP 4653155
A method (100) is disclosed of classifying elements in a region within a frame (410). The method (100) creates (145) a background model (150) of at least said region based on a statistical function applied to features of elements of said region in a set of frames. A mapping (130) of features to labels is also received. A difference measure comprising a plurality of difference values is then calculated based on features of elements in the region and features of the background model (150). The…
A method (100) is disclosed of classifying elements in a region within a frame (410). The method (100) creates (145) a background model (150) of at least said region based on a statistical function applied to features of elements of said region in a set of frames. A mapping (130) of features to labels is also received. A difference measure comprising a plurality of difference values is then calculated based on features of elements in the region and features of the background model (150). The method (100) then classifies (180) elements based upon the difference measure and the mapping (130).
Other inventors -
Honors & Awards
-
CVPR Best Paper Prize
The Conference on Computer Vision and Pattern Recognition (CVPR)
CVPR is the primary publication venue in Computer Vision, and the Best paper Prize is one of the most sought after awards in the field.
Recommendations received
1 person has recommended Anton
Join now to viewExplore more posts
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore More