Reinforcement Learning for Improving Object Detection

TL;DR

This paper introduces ObjectRL, a reinforcement learning algorithm that optimizes image pre-processing to enhance the performance of pre-trained object detection neural networks, addressing the mismatch between human perception and detector sensitivity.

Contribution

The paper presents a novel reinforcement learning approach, ObjectRL, for automatically selecting image pre-processing parameters to improve object detection accuracy.

Findings

ObjectRL improves detection performance on benchmark datasets.

The method outperforms traditional pre-processing techniques.

Optimized pre-processing leads to better detection accuracy.

Abstract

The performance of a trained object detection neural network depends a lot on the image quality. Generally, images are pre-processed before feeding them into the neural network and domain knowledge about the image dataset is used to choose the pre-processing techniques. In this paper, we introduce an algorithm called ObjectRL to choose the amount of a particular pre-processing to be applied to improve the object detection performances of pre-trained networks. The main motivation for ObjectRL is that an image which looks good to a human eye may not necessarily be the optimal one for a pre-trained object detector to detect objects.