Unsupervised learning from video to detect foreground objects in single images.
Ioana CroitoruSimion-Vlad BogolinMarius LeordeanuPublished in: CoRR (2017)
Keyphrases
- unsupervised learning
- foreground objects
- video sequences
- complex background
- moving camera
- input image
- three dimensional
- image features
- video surveillance
- dynamic scenes
- image retrieval
- image classification
- background subtraction
- video data
- video frames
- object recognition
- image annotation
- test images
- feature points
- supervised learning
- moving objects
- segmentation algorithm
- training data
- target object
- bounding box
- image sequences
- multiscale
- feature selection