We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: Cascaded Video Generation for Videos In-the-Wild

Abstract: Videos can be created by first outlining a global view of the scene and then adding local details. Inspired by this idea we propose a cascaded model for video generation which follows a coarse to fine approach. First our model generates a low resolution video, establishing the global scene structure, which is then refined by subsequent cascade levels operating at larger resolutions. We train each cascade level sequentially on partial views of the videos, which reduces the computational complexity of our model and makes it scalable to high-resolution videos with many frames. We empirically validate our approach on UCF101 and Kinetics-600, for which our model is competitive with the state-of-the-art. We further demonstrate the scaling capabilities of our model and train a three-level model on the BDD100K dataset which generates 256x256 pixels videos with 48 frames.
Comments: Accepted to the 26th International Conference on Pattern Recognition (ICPR 2022). arXiv admin note: substantial text overlap with arXiv:2106.02719
Subjects: Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
Cite as: arXiv:2206.00735 [cs.CV]
  (or arXiv:2206.00735v1 [cs.CV] for this version)

Submission history

From: Lluis Castrejon [view email]
[v1] Wed, 1 Jun 2022 19:50:50 GMT (1590kb,D)

Link back to: arXiv, form interface, contact.