Feedback neural network system, training method and device thereof and computer equipment

The invention discloses a feedback neural network system and a training method and device thereof, and computer equipment, the feedback neural network system adopts a sequence frame fn in a video stream as input, and outputs a feature ftn of a corresponding video sequence frame through forward networks net1 and net2 and splicing processing. According to the feedback neural network system disclosed by the invention, the feedback channel of the picture features exists, so that the features ftn of the sequence frames are fused with the information of the first n frames, and the problems of too low picture resolution and video jitter in the video stream can be solved. 本发明公开了一种反馈神经网络系统及其训练方法、装置、计算机设备，反馈神经网络系统采用视频流中序列帧fn作为输入，通过前向网络net1、net2及拼接处理输出对应视频序列帧的特征ftn。本发明公开的反馈神经网络系统中存在图片特征的反馈通道，使序列帧的特征ftn融合了前n帧的信息，能够解决视频流中图片分辨率过低以及视频抖动的问题。