{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,4]],"date-time":"2026-05-04T10:23:41Z","timestamp":1777890221711,"version":"3.51.4"},"reference-count":72,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,10,19]],"date-time":"2025-10-19T00:00:00Z","timestamp":1760832000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,10,19]]},"DOI":"10.1109\/iccv51701.2025.01574","type":"proceedings-article","created":{"date-parts":[[2026,4,29]],"date-time":"2026-04-29T19:45:49Z","timestamp":1777491949000},"page":"16948-16958","source":"Crossref","is-referenced-by-count":0,"title":["From Image to Video: An Empirical Study of Diffusion Representations"],"prefix":"10.1109","author":[{"given":"Pedro","family":"V\u00e9lez","sequence":"first","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Luisa F.","family":"Polan\u00eda","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Yi","family":"Yang","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Chuhan","family":"Zhang","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Rishabh","family":"Kabra","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Anurag","family":"Arnab","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Mehdi S. M.","family":"Sajjadi","sequence":"additional","affiliation":[{"name":"Google DeepMind"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01499"},{"key":"ref2","article-title":"Learning by reconstruction produces uninformative features for perception","author":"Balestriero","year":"2024","journal-title":"arXiv preprint"},{"key":"ref3","author":"Bao","year":"2024","journal-title":"Video diffusion models learn the structure of the dynamic world"},{"key":"ref4","article-title":"Label-efficient semantic segmentation with diffusion models","volume-title":"ICLR","author":"Baranchuk","year":"2022"},{"key":"ref5","article-title":"Revisiting feature prediction for learning visual representations from video","author":"Bardes","year":"2024","journal-title":"arXiv preprint"},{"key":"ref6","article-title":"Stable Video Diffusion: Scaling latent video diffusion models to large datasets","volume-title":"arXiv preprint","author":"Blattmann","year":"2023"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/3DV53792.2021.00027"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV48922.2021.00951"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.502"},{"key":"ref10","article-title":"A short note on the kinetics-700 human action dataset","author":"Carreira","year":"2019","journal-title":"arXiv preprint"},{"key":"ref11","article-title":"Scaling 4d representations","author":"Carreira","year":"2024","journal-title":"arXiv preprint"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00090"},{"key":"ref13","article-title":"Pali-3 vision language models: Smaller, faster, stronger","author":"Chen","year":"2023","journal-title":"arXiv preprint"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2571"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2017.261"},{"key":"ref16","article-title":"Depth map prediction from a single image using a multi-scale deep network","volume-title":"NeurIPS","author":"Eigen","year":"2014"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.01268"},{"key":"ref18","article-title":"Masked autoencoders as spatiotemporal learners","volume-title":"NeurIPS","author":"Feichtenhofer","year":"2022"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2026.3658965"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2017.622"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00373"},{"key":"ref22","article-title":"Photorealistic video generation with diffusion models","volume-title":"ECCV","author":"Gupta","year":"2024"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01553"},{"key":"ref24","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02153"},{"key":"ref25","article-title":"Denoising diffusion probabilistic models","volume-title":"NeurIPS","author":"Ho","year":"2020"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52734.2025.00193"},{"key":"ref27","article-title":"Elucidating the design space of diffusion-based generative models","volume-title":"NeurIPS","author":"Karras","year":"2022"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00210"},{"key":"ref29","article-title":"Sora: A review on background, technology, limitations, and opportunities of large vision models","author":"Liu","year":"2024","journal-title":"arXiv preprint"},{"key":"ref30","article-title":"Diffusion hyperfeatures: Searching through time and space for semantic correspondence","volume-title":"NeurIPS","author":"Luo","year":"2023"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2015.7299155"},{"key":"ref32","article-title":"Lexicon3d: Probing visual foundation models for complex 3d scene understanding","volume-title":"The Thirty-eighth Annual Conference on Neural Information Processing Systems","author":"Man","year":"2024"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58452-8_24"},{"key":"ref34","article-title":"Diffusion models beat GANs on image classification","author":"Mukhopadhyay","year":"2023","journal-title":"arXiv preprint"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00951"},{"key":"ref36","article-title":"DINOv2: Learning robust visual features without supervision","volume-title":"TMLR","author":"Oquab","year":"2024"},{"key":"ref37","article-title":"Self-supervised video pretraining yields robust and more human-aligned visual representations","volume-title":"NeurIPS","author":"Parthasarathy","year":"2023"},{"key":"ref38","article-title":"Perception Test: A diagnostic benchmark for multimodal video models","volume-title":"NeurIPS","author":"Patraucean","year":"2024"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00387"},{"key":"ref40","article-title":"The 2017 DAVIS challenge on video object segmentation","author":"Pont-Tuset","year":"2017","journal-title":"arXiv:1704.00675"},{"key":"ref41","article-title":"Learning transferable visual models from natural language supervision","volume-title":"ICML","author":"Radford","year":"2021"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"ref43","article-title":"UNet: Convolutional networks for biomedical image segmentation","volume-title":"MICCAI","author":"Ronneberger","year":"2015"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-015-0816-y"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00613"},{"key":"ref46","article-title":"Where\u2019s Waldo: Diffusion features for personalized segmentation and retrieval","volume-title":"NeurIPS","author":"Samuel","year":"2024"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1109\/WACV48630.2021.00058"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2024.3509619"},{"key":"ref49","article-title":"Deep unsupervised learning using nonequilibrium thermodynamics","volume-title":"ICML","author":"Sohl-Dickstein","year":"2015"},{"key":"ref50","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR42600.2020.00252"},{"key":"ref51","article-title":"Emergent correspondence from image diffusion","volume-title":"NeurIPS","author":"Tang","year":"2023"},{"key":"ref52","article-title":"VideoMAE: Masked autoencoders are data-efficient learners for self-supervised video pre-training","volume-title":"NeurIPS","author":"Tong","year":"2022"},{"key":"ref53","article-title":"Towards accurate generative models of video: A new metric & challenges","author":"Unterthiner","year":"2018","journal-title":"arXiv preprint"},{"key":"ref54","article-title":"Neural discrete representation learning","volume-title":"NeurIPS","author":"Oord","year":"2017"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2018.00914"},{"key":"ref56","article-title":"Moving off-the-grid: Scene-grounded video representations","volume-title":"NeurIPS","author":"van Steenkiste","year":"2024"},{"key":"ref57","article-title":"Attention is all you need","volume-title":"NeurIPS","author":"Vaswani","year":"2017"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01398"},{"key":"ref59","article-title":"Controlling space and time with diffusion models","author":"Watson","year":"2024","journal-title":"arXiv preprint"},{"key":"ref60","article-title":"GenRec: Unifying video generation and recognition with diffusion models","volume-title":"NeurIPS","author":"Weng","year":"2024"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01448"},{"key":"ref62","article-title":"Video diffusion models are training-free motion interpreter and controller","author":"Xiao","year":"2024","journal-title":"arXiv preprint"},{"key":"ref63","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.00289"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.01736"},{"key":"ref65","article-title":"Language model beats diffusion - tokenizer is key to visual generation","volume-title":"ICLR","author":"Yu","year":"2024"},{"key":"ref66","article-title":"A tale of two features: Stable diffusion complements DINO for zero-shot semantic correspondence","volume-title":"NeurIPS","author":"Zhang","year":"2024"},{"key":"ref67","article-title":"A survey of diffusion based image generation models: Issues and their solutions","author":"Zhang","year":"2023","journal-title":"arXiv preprint"},{"key":"ref68","article-title":"VideoPrism: A foundational visual encoder for video understanding","volume-title":"ICML","author":"Zhao","year":"2024"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV51070.2023.00527"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2017.2723009"},{"key":"ref71","doi-asserted-by":"publisher","DOI":"10.1145\/3197517.3201323"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73254-6_26"}],"event":{"name":"2025 IEEE\/CVF International Conference on Computer Vision (ICCV)","location":"Honolulu, HI, USA","start":{"date-parts":[[2025,10,19]]},"end":{"date-parts":[[2025,10,25]]}},"container-title":["2025 IEEE\/CVF International Conference on Computer Vision (ICCV)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11443115\/11443287\/11445686.pdf?arnumber=11445686","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,1]],"date-time":"2026-05-01T05:32:28Z","timestamp":1777613548000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11445686\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,19]]},"references-count":72,"URL":"https:\/\/doi.org\/10.1109\/iccv51701.2025.01574","relation":{},"subject":[],"published":{"date-parts":[[2025,10,19]]}}}