Neural Video Fields Editing
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2023
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917609113387008 |
|---|---|
| author | Yang, Shuzhou Mou, Chong Yu, Jiwen Wang, Yuhan Meng, Xiandong Zhang, Jian |
| author_facet | Yang, Shuzhou Mou, Chong Yu, Jiwen Wang, Yuhan Meng, Xiandong Zhang, Jian |
| contents | Diffusion models have revolutionized text-driven video editing. However, applying these methods to real-world editing encounters two significant challenges: (1) the rapid increase in GPU memory demand as the number of frames grows, and (2) the inter-frame inconsistency in edited videos. To this end, we propose NVEdit, a novel text-driven video editing framework designed to mitigate memory overhead and improve consistent editing for real-world long videos. Specifically, we construct a neural video field, powered by tri-plane and sparse grid, to enable encoding long videos with hundreds of frames in a memory-efficient manner. Next, we update the video field through off-the-shelf Text-to-Image (T2I) models to impart text-driven editing effects. A progressive optimization strategy is developed to preserve original temporal priors. Importantly, both the neural video field and T2I model are adaptable and replaceable, thus inspiring future research. Experiments demonstrate the ability of our approach to edit hundreds of frames with impressive inter-frame consistency. Our project is available at: https://nvedit.github.io/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2312_08882 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | Neural Video Fields Editing Yang, Shuzhou Mou, Chong Yu, Jiwen Wang, Yuhan Meng, Xiandong Zhang, Jian Computer Vision and Pattern Recognition Diffusion models have revolutionized text-driven video editing. However, applying these methods to real-world editing encounters two significant challenges: (1) the rapid increase in GPU memory demand as the number of frames grows, and (2) the inter-frame inconsistency in edited videos. To this end, we propose NVEdit, a novel text-driven video editing framework designed to mitigate memory overhead and improve consistent editing for real-world long videos. Specifically, we construct a neural video field, powered by tri-plane and sparse grid, to enable encoding long videos with hundreds of frames in a memory-efficient manner. Next, we update the video field through off-the-shelf Text-to-Image (T2I) models to impart text-driven editing effects. A progressive optimization strategy is developed to preserve original temporal priors. Importantly, both the neural video field and T2I model are adaptable and replaceable, thus inspiring future research. Experiments demonstrate the ability of our approach to edit hundreds of frames with impressive inter-frame consistency. Our project is available at: https://nvedit.github.io/. |
| title | Neural Video Fields Editing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2312.08882 |