MIST: Multiple Instance Spatial Transformer Network
This addresses the challenge of unsupervised object detection for computer vision applications, offering a novel training approach for such problems.
The paper tackles the problem of detecting multiple object instances in images without location supervision by proposing a network that learns to extract top-K patches and uses them for reconstruction or classification, achieving state-of-the-art performance in localization tasks.
We propose a deep network that can be trained to tackle image reconstruction and classification problems that involve detection of multiple object instances, without any supervision regarding their whereabouts. The network learns to extract the most significant top-K patches, and feeds these patches to a task-specific network -- e.g., auto-encoder or classifier -- to solve a domain specific problem. The challenge in training such a network is the non-differentiable top-K selection process. To address this issue, we lift the training optimization problem by treating the result of top-K selection as a slack variable, resulting in a simple, yet effective, multi-stage training. Our method is able to learn to detect recurrent structures in the training dataset by learning to reconstruct images. It can also learn to localize structures when only knowledge on the occurrence of the object is provided, and in doing so it outperforms the state-of-the-art.