Hierarchical Object Detection with Deep Reinforcement Learning

TL;DR

This paper introduces a hierarchical object detection method using deep reinforcement learning to focus on informative image regions, comparing candidate strategies and feature extraction methods, achieving improved detection efficiency.

Contribution

The work presents a novel reinforcement learning-based approach for hierarchical object detection, including strategies for region proposal and feature extraction, with experimental validation.

Findings

Overlapping candidate proposal strategy performs better.

Feature extraction from the whole image is more effective than cropping.

Reinforcement learning reduces the number of region proposals needed.

Abstract

We present a method for performing hierarchical object detection in images guided by a deep reinforcement learning agent. The key idea is to focus on those parts of the image that contain richer information and zoom on them. We train an intelligent agent that, given an image window, is capable of deciding where to focus the attention among five different predefined region candidates (smaller windows). This procedure is iterated providing a hierarchical image analysis.We compare two different candidate proposal strategies to guide the object search: with and without overlap. Moreover, our work compares two different strategies to extract features from a convolutional neural network for each region proposal: a first one that computes new feature maps for each region proposal, and a second one that computes the feature maps for the whole image to later generate crops for each region proposal. Experiments indicate better results for the overlapping candidate proposal strategy and a loss of performance for the cropped image features due to the loss of spatial resolution. We argue that, while this loss seems unavoidable when working with large amounts of object candidates, the much more reduced amount of region proposals generated by our reinforcement learning agent allows considering to extract features for each location without sharing convolutional computation among regions.