YOLO-Owl: An Occlusion Aware Detector for Low Illuminance Environment

Object detection algorithms can be roughly divided into two categories, two-stage detectors like Fast RCNN and one-stage detectors like YOLO. We propose an object detection model that works well in low illuminance environment. We design the MRFM and FEID module for low illuminance targets, as well as a new training model method for low illumination target that allows the model to converge stably and quickly. In addition, we propose an occlusion-aware attention module MPCM to deal with target occlusion in low-light environment. We expand the ExDark dataset and run experiments on the expanded dataset. The results show that the performance of our proposed model is better than YOLOv5 and YOLOX.