For the production process to realize rapid and accurate detection of the quality of the film on the skin of the Yuba, to realize the automated peeling of the skin of the Yuba skin. Development of conjunctival quality determination criteria based on changes in surface characteristics of Yuba skin, According to the time-varying characteristics of Yuba skin quality, an improved YOLOv5s method was proposed for the identification and detection of Yuba skin. This method makes the bottleneck in the C3 module of Backbone in YOLOv5s be replaced with CBAM (Convolutional block attention module) attention, which enhances the network feature extraction capability and localizes the target more effectively. The PANet at the Neck end is replaced with BiFPN, which enhances the feature fusion capability by a weighted bidirectional feature pyramid network, thus improving the recognition efficiency and accuracy. Finally, the loss function is replaced with the WioU loss function. The experimental results show that the recall, accuracy, and mean average precision mAP (Mean AveragePrecision) of the improved YOLOv5s model are 89.6%, 94.6%, and 92.3%, respectively. The improved model can more accurately identify the Yuba skin that meets the conjunctiva standard, At the same time, due to its advantages of high detection accuracy and fast detection speed, it also provides technical support for improving the automated and continuous production of high-quality Yuba skin.