Tongue-Segmentation-and-classification

😁This project utilizes the SAM algorithm for segmenting tongues in images, followed by classification using ViT.

Prediction

1. Place the tongue images you want to recognize in the data/orin folder.
2. Run the predict.py script. You can choose one of the following folders as the inference image path:
   • To use the cropped tongue bounding box images, set the variable cropped_img_path to ./data/cropped.
   • To use the segmented tongue images, set the variable seg_img_path to ./data/seg.

Classification Json

• The current classification is based on tongue color and tongue coating status. Details can be found in the 2 class_indices.json files.

Fine-tuning

If you need to perform fine-tuning, you can use the code provided in train.py.

Pretrained Models

• Download the tongue segmentation model (tonguesam.pth) from the following link: Download here (Place tonguesam.pth in the ./pretrained_model/ folder).
• Download the ViT model (ViT.pth) from the following link: Download here (Place ViT.pth in the ./weight/ folder).

License

This project is licensed under the MIT License. See the LICENSE file for details.