Bringing Image Scene Structure to Video via Frame-Clip Consistency of Object Tokens.
Elad Ben-AvrahamRoei HerzigKarttikeya MangalamAmir BarAnna RohrbachLeonid KarlinskyTrevor DarrellAmir GlobersonPublished in: NeurIPS (2022)
Keyphrases
- key frames
- scene structure
- video frames
- image motion
- single image
- image measurements
- input image
- multiple objects
- ego motion
- multiscale
- target object
- video sequences
- region of interest
- image segmentation
- image features
- position and orientation
- image sequences
- feature points
- d objects
- closed form
- bundle adjustment
- d scene
- multi frame
- structure from motion
- camera motion
- feature correspondences