Improved YOLOv5s for Object Detection

The utilization of deep learning technology for object detection offers the notable benefit of attaining high levels of accuracy. The level of accuracy attained is contingent upon the duration of processing time required by the deep learning technology. The algorithm for object detection is commonly referred to as You Only Look Once (YOLO). This study presents a novel object detection algorithm utilizing the Yolov5 framework, coupled with a substantial dataset. The improvement of YOLOv5 commenced with the incorporation of the Spatial Pyramid Pooling Fast module (SPPF) into the yolov5s neck network to extract multiple scale image features. Then, the depth of the backbone SPPF module has been raised from 1024 to 2048 to improve accuracy. The modified YOLOV5 model was utilized for object detection, resulting in the extraction of object information categorized into seven distinct classes. The achieved results include a mean average precision (mAP) of 0.821 at an intersection over union (IoU) threshold of 0.5, and a mAP of 0.386 within the IoU range of 0.5 to 0.95, with a precision value of 0.761. Additionally, a recall value of 0.769 was obtained. The application of our research framework has demonstrated its effectiveness in enhancing the performance of object detection algorithms.