Attend Refine Repeat: Active Box Proposal Generation via In-Out Localization

The problem of computing category agnostic bounding box proposals is utilized a core component in many computer vision tasks and thus has lately attracted lot of attention. In this work we propose a new approach to tackle this that is based on an active strategy for generating box proposals that from a set of seed boxes, which are uniformly distributed on the image, then progressively moves its attention on the promising image areas where is more likely to discover well localized bounding box proposals. We call approach AttractioNet and a core component of it is a CNN-based category object location refinement module that is capable of yielding accurate robust bounding box predictions regardless of the object category. We extensively evaluate our AttractioNet approach on several image datasets (i. e. COCO, PASCAL, ImageNet detection and NYU-Depth V2 datasets) reporting on of them state-of-the-art results that surpass the previous work in the by a significant margin and also providing strong empirical evidence that approach is capable to generalize to unseen categories. Furthermore, we our AttractioNet proposals in the context of the object detection task a VGG16-Net based detector and the achieved detection performance on COCO to significantly surpass all other VGG16-Net based detectors while even competitive with a heavily tuned ResNet-101 based detector. Code as well box proposals computed for several datasets are available at: : : //github. com/gidariss/AttractioNet.