Learning to Segment Moving Objects in Videos
This work addresses video segmentation for computer vision applications, offering incremental improvements in detection rates and benchmark performance.
The paper tackles the problem of segmenting moving objects in videos by ranking spatio-temporal segment proposals based on 'moving objectness', achieving a 7% increase in detection rate over previous state-of-the-art static proposal methods and outperforming previous segmentation methods in benchmarks.
We segment moving objects in videos by ranking spatio-temporal segment proposals according to "moving objectness": how likely they are to contain a moving object. In each video frame, we compute segment proposals using multiple figure-ground segmentations on per frame motion boundaries. We rank them with a Moving Objectness Detector trained on image and motion fields to detect moving objects and discard over/under segmentations or background parts of the scene. We extend the top ranked segments into spatio-temporal tubes using random walkers on motion affinities of dense point trajectories. Our final tube ranking consistently outperforms previous segmentation methods in the two largest video segmentation benchmarks currently available, for any number of proposals. Further, our per frame moving object proposals increase the detection rate up to 7\% over previous state-of-the-art static proposal methods.