Learning Features by Watching Objects Move
This work addresses the problem of learning effective visual representations without labeled data for computer vision researchers, offering a more intuitive approach than existing pretext tasks, though it is incremental in building on motion-based cues.
The paper tackles unsupervised visual feature learning by using motion-based segmentation from videos as pseudo ground truth to train a convolutional network for object segmentation from single frames, resulting in a representation that significantly outperforms previous unsupervised methods in object detection transfer learning, particularly with scarce training data.
This paper presents a novel yet intuitive approach to unsupervised feature learning. Inspired by the human visual system, we explore whether low-level motion-based grouping cues can be used to learn an effective visual representation. Specifically, we use unsupervised motion-based segmentation on videos to obtain segments, which we use as 'pseudo ground truth' to train a convolutional network to segment objects from a single frame. Given the extensive evidence that motion plays a key role in the development of the human visual system, we hope that this straightforward approach to unsupervised learning will be more effective than cleverly designed 'pretext' tasks studied in the literature. Indeed, our extensive experiments show that this is the case. When used for transfer learning on object detection, our representation significantly outperforms previous unsupervised approaches across multiple settings, especially when training data for the target task is scarce.