Due to human exploitation of underground mineral resources, mining area collapse disasters occur frequently, and large area of ground collapse forms a subsidence funnel. The automatic recognition of subsidence funnel in a large area is an urgent need for the prevention and treatment of geological disasters in mining areas. In this paper, six advanced deep learning target detection networks are used to train and test models on 117 real interferograms obtained using D-InSAR technology. The experimental results show that it is feasible to use the depth learning target detection method to automatically detect the mining subsidence funnel on InSAR interferograms. Among them, the YOLO-V5 model with CSPDarknet shows the best performance with the mAP of 89.97% and the F1-Score of 0.83.