HOUSE IDENTIFICATION IN LANDSLIDE AREA BASED ON YOLOV3-HA

Aimed at the problem of large size and low efficiency of the YOLOv3 model, YOLOv3-HA, an improved method for identifying houses in landslide areas based on YOLOv3, is proposed. HetConv was used to replace the conventional convolution kernel, the CBAM module and the pyramid pooling structure were introduced to improve the performance of the model, and a more accurate EIoU was used as the frame regression loss. The experimental results on the landslide house dataset show that compared with the original model, the size of the model and FLOPs are reduced by about 70%, detection efficiency are improved by 20%, and the detection accuracy is increased by 4.27 percentage points, which verifies the effectiveness of the lightweight algorithm.