Workpiece Recognition and Localization Based on Improved YOLOv7-Tiny Channel Pruning

At present, to address the problem of difficult detection of industrial parts and complex model with large computation, a workpiece recognition and detection algorithm based on improved YOLOv7-Tiny channel pruning is proposed. Firstly, the algorithm integrates the ECA module, an efficient channel attention mechanism, into the backbone network to better retain the information in transmission, so that the data features are more fully retained in the process of transmission, and thus the model can be optimized and the detection accuracy of the model can be improved; secondly, the lightweight up-sampling operator CARAFE is invoked to replace the nearest-neighbor difference of the original up-sampling in the structure of the PAFPN in order to improve the feature fusion ability of the model; finally, the improved model is subjected to channel pruning to reduce the number of parameters and computation of the model to further optimize the model. The experimental results show that our method achieves 81.5% recognition accuracy on the self-constructed dataset, and the number of parameters and computation amount decrease by 75.6% and 74.0%, respectively. The effectiveness of the method is proved.