A model that can detect face emotion like negative ,positive and normal based on Face-emotion-yolov5 with the link https://github.com/SalehAliAlyahri/Face-emotion-yolov5 I use the model in Face-emotion-yolov5 project for transfering learning Based on transfer learning, I used three labels: normal, negative, and positive. I believe that human emotions are not just sad and happy, sometimes there may also be anger, and these emotions are all negative. Therefore, I replaced happy with positive and sad with negative. From the results, the recognition accuracy of normal is the highest, but sometimes positive and negative cannot be distinguished. I think this is due to the small size of the dataset and the similarity of the photos of positive and negative emotions, causing overfitting. I still change the detect.py in yolov5 so that it can be packaged into an exe file better
the accuracy is not ideal,you can still use this model in this project for transfering learning to achieve your purpose you want.