Feature-Driven Super-Resolution for Object Detection
This addresses the issue of poor detection accuracy from low-resolution images for computer vision applications, but it is incremental as it builds on existing SR and detection methods.
The paper tackles the problem of low-resolution images degrading object detection performance by proposing a feature-driven super-resolution method that uses detector backbone features to guide reconstruction, resulting in improved mAP on MS COCO and VOC2007 datasets compared to state-of-the-art SR algorithms.
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.