Reinforcement Learning for Visual Object Detection

• Published in: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Vol. 2016-January; pp. 2894 - 2902
• Main Authors: Mathe, Stefan, Pirinen, Aleksis, Sminchisescu, Cristian
• Format: Conference Proceeding Book Chapter
• Language: English
• Published: IEEE 01.06.2016
• Subjects: Computational efficiency; Computational modeling; Computer and Information Science; Computer and Information Sciences; Computer graphics and computer vision; Computer Vision and Robotics (Autonomous Systems); Data- och informationsvetenskap (Datateknik); Datorgrafik och datorseende; Datorseende och robotik (autonoma system); Detectors; Feature extraction; History; Natural Sciences; Naturvetenskap; Object detection; Visualization
• DOI: 10.1109/CVPR.2016.316
• ISBN: 9781467388511; 1467388513

One of the most widely used strategies for visual object detection is based on exhaustive spatial hypothesis search. While methods like sliding windows have been successful and effective for many years, they are still brute-force, independent of the image content and the visual category being searched. In this paper we present principled sequential models that accumulate evidence collected at a small set of image locations in order to detect visual objects effectively. By formulating sequential search as reinforcement learning of the search policy (including the stopping condition), our fully trainable model can explicitly balance for each class, specifically, the conflicting goals of exploration - sampling more image regions for better accuracy -, and exploitation - stopping the search efficiently when sufficiently confident about the target's location. The methodology is general and applicable to any detector response function. We report encouraging results in the PASCAL VOC 2012 object detection test set showing that the proposed methodology achieves almost two orders of magnitude speed-up over sliding window methods.