Infinite Gaze Generation for Videos with Autoregressive Diffusion
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866917362332073984 |
|---|---|
| author | Kang, Jenna Groth, Colin Wu, Tong Torrens, Finley Sangkloy, Patsorn Wetzstein, Gordon Sun, Qi |
| author_facet | Kang, Jenna Groth, Colin Wu, Tong Torrens, Finley Sangkloy, Patsorn Wetzstein, Gordon Sun, Qi |
| contents | Predicting human gaze in video is fundamental to advancing scene understanding and multimodal interaction. While traditional saliency maps provide spatial probability distributions and scanpaths offer ordered fixations, both abstractions often collapse the fine-grained temporal dynamics of raw gaze. Furthermore, existing models are typically constrained to short-term windows ($\approx$ 3-5s), failing to capture the long-range behavioral dependencies inherent in real-world content. We present a generative framework for infinite-horizon raw gaze prediction in videos of arbitrary length. By leveraging an autoregressive diffusion model, we synthesize gaze trajectories characterized by continuous spatial coordinates and high-resolution timestamps. Our model is conditioned on a saliency-aware visual latent space. Quantitative and qualitative evaluations demonstrate that our approach significantly outperforms existing approaches in long-range spatio-temporal accuracy and trajectory realism. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_24938 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Infinite Gaze Generation for Videos with Autoregressive Diffusion Kang, Jenna Groth, Colin Wu, Tong Torrens, Finley Sangkloy, Patsorn Wetzstein, Gordon Sun, Qi Computer Vision and Pattern Recognition Predicting human gaze in video is fundamental to advancing scene understanding and multimodal interaction. While traditional saliency maps provide spatial probability distributions and scanpaths offer ordered fixations, both abstractions often collapse the fine-grained temporal dynamics of raw gaze. Furthermore, existing models are typically constrained to short-term windows ($\approx$ 3-5s), failing to capture the long-range behavioral dependencies inherent in real-world content. We present a generative framework for infinite-horizon raw gaze prediction in videos of arbitrary length. By leveraging an autoregressive diffusion model, we synthesize gaze trajectories characterized by continuous spatial coordinates and high-resolution timestamps. Our model is conditioned on a saliency-aware visual latent space. Quantitative and qualitative evaluations demonstrate that our approach significantly outperforms existing approaches in long-range spatio-temporal accuracy and trajectory realism. |
| title | Infinite Gaze Generation for Videos with Autoregressive Diffusion |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.24938 |