Feature-Driven Super-Resolution for Object Detection

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 high-resolution (HR) image reconstruction. Then, with the aligned features, FDSR updates SR parameters for better detection performance. Comparing with some state-of-the-art SR algorithms with scale factor 4, FDSR outperforms the detection performance mAP on MS COCO validation and VOC2007 test datasets, and with good generalization to other detection networks.