Semantic communication method, device and equipment for binocular three-dimensional target recognition

The invention discloses a semantic communication method, device and equipment for binocular three-dimensional target detection, and the method comprises the steps: inputting obtained left and right visual pictures into a trained neural network, and extracting key region semantic information and global semantic information; respectively carrying out compression coding on the extracted semantic information at different compression ratios to obtain a compressed key region and global semantic information; respectively carrying out channel coding on the compressed semantic information to obtain channel coding information, and transmitting the channel coding information to a receiver through a channel; decoding the received channel coding information, and performing semantic recovery on the key region semantic information and the global semantic information which are still in a compressed state after decoding to obtain the key region semantic information and the global semantic information; and fusing the recovered