Feature-Driven Super-Resolution for Object Detection

TL;DR

This paper introduces a feature-driven super-resolution method that enhances low-resolution images specifically for better object detection performance, outperforming existing SR algorithms on standard datasets.

Contribution

The proposed FDSR method uniquely incorporates detector features to guide super-resolution, improving detection accuracy on low-resolution images.

Findings

FDSR outperforms state-of-the-art SR methods in detection mAP.

FDSR generalizes well across different detection networks.

FDSR effectively enhances detection performance on MS COCO and VOC datasets.

Abstract

Although some convolutional neural networks (CNNs) based super-resolution (SR) algorithms yield good visual performances on single images recently. Most of them focus on perfect perceptual quality but ignore specific needs of subsequent detection task. This paper proposes a simple but powerful feature-driven super-resolution (FDSR) to improve the detection performance of low-resolution (LR) images. First, the proposed method uses feature-domain prior which extracts from an existing detector backbone to guide the HR image reconstruction. Then, with the aligned features, FDSR update SR parameters for better detection performance. Comparing with some state-of-the-art SR algorithms with 4$\times$ scale factor, FDSR outperforms the detection performance mAP on MS COCO validation, VOC2007 databases with good generalization to other detection networks.