Task-Driven Super Resolution: Object Detection in Low-Resolution Images

We consider how image super-resolution (SR) can contribute to an object detection task in low-resolution images. Intuitively, SR gives a positive impact on the object detection task. While several previous works demonstrated that this intuition is correct, SR and detector are optimized independently in these works. This paper analyze a framework to train a deep neural network where the SR sub-network explicitly incorporates a detection loss in its training objective, via a tradeoff with a traditional detection loss. This end-to-end training procedure allows us to train SR preprocessing for any differentiable detector. We demonstrate extensive experiments that show our task-driven SR consistently and significantly improves the accuracy of an object detector on low-resolution images from COCO and PASCAL VOC data set for a variety of conditions and scaling factors.