Research on intelligent defect detection methods in medical balloons
In response to issues such as low efficiency and quality variability in manual detection of defects in medical balloons due to human experience and subjective factors,a lightweight detection method based on improving the YOLOv5s algorithm is proposed.To achieve better robustness,this study autonomously constructed a dataset based on defect occurrences during medical balloon production.Firstly,the backbone network of YOLOv5s was replaced with the FasterNet network structure to significantly lighten the network while maintaining detection accuracy and improving detection speed.Secondly,the Content-Aware ReAssembly of FEatures(CARAFE)upsampling operator was introduced to increase the receptive field and enhance the reconstruction quality of feature maps,thereby improving the model's detection accuracy.Lastly,the Coordinate Attention(CA)mechanism was introduced in the feature extraction stage to enhance the network's ability to detect small target defects.Testing on the constructed dataset of medical balloon defects showed that compared to the original YOLOv5s algorithm,the proposed algorithm achieved an average precision mean average precision(mAP)improvement of 1.7%,reduced floating-point operations per second(FLOPS)by 8.4×109,decreased weight size by 86.9%to 7.5MB,and increased frame rate by 12.7 frames per second to 71.5 frames/s,significantly lightening the overall model.
medical balloondefect detectionYOLOv5FasterNetCACARAFE
NETL
NSTL
万方数据
