Vertical and Horizontal Network for Small Object Detection in Sports Videos
Xiao Han, Yongbin Wang, Qi Wang, Nenghuan Zhang, Chenming Liu · 2024
Detecting small objects is often impeded by blurriness and low resolution, which poses substantial challenges for accurately detecting and localizing such objects. Balls and players in sports competition videos are small objects more difficult to detect because of motion blur. Balls and players in sports game videos are small objects that are more difficult to detect due to motion blur. To tackle this challenge, we propose a small object detection backbone network named VHNet based on vertical and horizontal information flow in sports scenarios. In order to verify the effectiveness of our backbone network, we manually annotated a small ball detection data set and conducted comparative experiments with the SOTA small object detection method. The results demonstrate the superiority of our method in sports video detection tasks.