Detector Discovery in the Wild: Joint Multiple Instance and Representation Learning
This work addresses the challenge of reducing annotation costs for object detection in computer vision, though it appears incremental as it builds on existing weak-label learning and transfer learning approaches.
The paper tackles the problem of learning object detectors from weakly-labeled data by proposing a model that jointly trains representations and detectors using both weak and strong labels, and it outperforms previous methods on ImageNet-200 detection with weak labels.
We develop methods for detector learning which exploit joint training over both weak and strong labels and which transfer learned perceptual representations from strongly-labeled auxiliary tasks. Previous methods for weak-label learning often learn detector models independently using latent variable optimization, but fail to share deep representation knowledge across classes and usually require strong initialization. Other previous methods transfer deep representations from domains with strong labels to those with only weak labels, but do not optimize over individual latent boxes, and thus may miss specific salient structures for a particular category. We propose a model that subsumes these previous approaches, and simultaneously trains a representation and detectors for categories with either weak or strong labels present. We provide a novel formulation of a joint multiple instance learning method that includes examples from classification-style data when available, and also performs domain transfer learning to improve the underlying detector representation. Our model outperforms known methods on ImageNet-200 detection with weak labels.