Static Video Compression’s Influence on Neural Network Performance

The concept of action recognition in smart security heavily relies on deep learning and artificial intelligence to make predictions about actions humans. To draw appropriate conclusions from these hypotheses, a large amount information is required. data question are often video feed, there direct relationship between increased volume more-precise decision-making. We seek determine how far static can be compressed before the neural network’s capacity predict lost. find this, videos by lowering bitrate using FFMPEG. In parallel, convolutional network model trained recognise tested until fails observed videos. results reveal that compression has no linear with performance.