Abstract: Aiming at the problem of coal gangue sorting, a target detection algorithm based on deep learning is proposed to replace manual gangue sorting. At the same time, a coal gangue image data expansion method based on deep convolution generation countermeasure network is proposed. Through experimental analysis, it is found that the improved yolov4 algorithm using k-means initial anchor frame parameters and adding a detection channel can overcome the defect of missing detection of small coal gangue, At the same time, the detection accuracy of the target detection model is improved.