Plastic Part Surface Defect Detection Based on Improved YOLOv7

To address the issues of low efficiency and poor accuracy in plastic part surface defect detection models, this paper proposes an improved YOLOv7-based detection algorithm, YOLOv7-TBC. First, a T-shaped convolutional module is introduced into the backbone network to enhance feature extraction capability and detection speed. Then, a weighted bidirectional feature pyramid network is applied at the neck to improve feature fusion efficiency and capacity. Finally, a convolutional block attention module is embedded to enhance the model's performance in detecting small objects. Experimental results show that the proposed algorithm achieves a mean average precision (mAP) of 97.4% on the plastic part dataset, representing a 1.5% improvement over the original model. Additionally, model parameters are reduced by 21.3%, and computational load is decreased by 17.4<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, significantly improving detection speed and accuracy, effectively meeting industrial defect detection requirements.