Is ImageNet worth 1 video? Learning strong image encoders from 1 long unlabelled video.
Shashanka VenkataramananMamshad Nayeem RizveJoão CarreiraYuki M. AsanoYannis AvrithisPublished in: ICLR (2024)
Keyphrases
- video data
- video sequences
- image data
- multimedia
- video content
- video images
- weakly labeled
- image frames
- learning process
- image collections
- learning algorithm
- image features
- video clips
- multiscale
- video files
- visual data
- key frames
- video frames
- input image
- single image
- video streams
- real time video
- static images
- video analysis
- image classification
- video retrieval
- test images
- image regions
- active learning
- multimedia data
- high resolution
- image retrieval
- image segmentation
- segmentation method
- visual concepts
- object motion
- compressed video
- temporal continuity