Detection of Local Lesions in Tongue Recognition Based on Improved Faster R-CNN

Tongue diagnosis is a crucial part of the four diagnostic methods of traditional Chinese medicine (TCM). The detection of local lesions in tongue recognition can help to identify disease progression and body constitution of the patient. In this study, we propose an improved faster region convolutional neural network (R-CNN) method based on residual network (ResNet50) model and attention mechanism. Firstly, taking into account the increase of missing tongue information with the rise of the depth of network, we propose to adapt the ResNet50 model in the region proposal network (RPN) to keep more original tongue information. Secondly, attention mechanism is used in the network to emphasize the local lesion information and restrain the background tongue information. The experimental results show that the proposed method allows to improve significantly the precision rate compared to the original Faster R-CNN.