Performance Evaluation of Detection Algorithm Based on Improved YOLOv1

In order to improve the detection performance of the YOLOv1 model for images, this paper has made improvements to the YOLOv1 model, mainly including the following aspects:(1) Replace YOLOv1's GoogLeNet style backbone network with ResNet network.(2) After the backbone network, select the Spatial Pyramid Pooling (SPP) module and improve it.(3) The decoupling detection head is used as the detection head of YOLOV1, consisting of category branch and regression branch, to extract category features and position features respectively.(4) Modify the prediction layer by using a 1x1 convolutional layer to make predictions on the feature map.(5) Modify YOLO's loss function, including confidence loss, category loss, and loss of bounding box position parameters. Finally, testing was conducted on the VOC2007 dataset, and the results showed that the improved model achieved higher performance.