基于注意力机制的足球视频目标检测

Soccer video object detection based on deep learning with attention mechanism

足球运动在全世界范围内拥有广泛的受众和巨大的市场,利用计算机视觉技术对足球比赛视频进行目标检测,自动识别出球员、足球的位置,可以为进一步的跟踪提供良好的基础,对球队、对转播都有很大的帮助。本文提出了基于注意力机制的神经网络足球视频目标检测方案,通过搭建使用注意力机制的深度神经网络,并在足球相关数据集上进行训练,实现了对目标较为完整、准确的检测,为进一步的跟踪提供了帮助。

Football has a wide audience and a huge market all over the world.Computer vision technology is used to detect the objects in football game videos and automatically identify the positions of players and football,which can provide a good foundation for further tracking task.It is of great help to broadcasting.This paper proposes a football video object detection scheme based on neural network using the attention mechanism:by building a deep neural network using the attention mechanism and training on football-related dataset,a more complete and accurate detection of the object is achieved and may help further tracking tasks.