Mark-CHAE/SM-Tongue-Public-Original512
SM-Tongue Public Original512 This dataset release contains 2,155 anonymized 512×512 tongue images with expert segmentation masks and overlay visualizations for reproducible tongue segmentation research. Download Download the archive from this repository: SM-Tongue-Public-Original512.tar.gz SM-Tongue-Public-Original512.tar.gz.sha256 Verify integrity: sha256sum -c SM-Tongue-Public-Original512.tar.gz.sha256 Extract: tar -xzf SM-Tongue-Public-Original512.tar.gz… See the full description on the dataset page: https://huggingface.co/datasets/Mark-CHAE/SM-Tongue-Public-Original512.
SM-Tongue Public Original512
This dataset release contains 2,155 anonymized 512×512 tongue images with expert segmentation masks and overlay visualizations for reproducible tongue segmentation research.
Download
Download the archive from this repository:
SM-Tongue-Public-Original512.tar.gzSM-Tongue-Public-Original512.tar.gz.sha256
Verify integrity:
sha256sum -c SM-Tongue-Public-Original512.tar.gz.sha256Extract:
tar -xzf SM-Tongue-Public-Original512.tar.gzContents
After extraction, the folder SM-Tongue-2155-anonymized/ contains:
images/: 2,155 input images, named000001.pngto002155.pngmasks/: 2,155 binary segmentation masks with matching namesoverlays/: 2,155 overlay visualizations with matching namesmanifest.csv: public image/mask/overlay path manifestmemory-sm_tongue/: 20 anonymized memory-bank examples used by the released Memory-SAM default settingREADME.mdandDATASET_META.json: dataset metadata
All images, masks, and overlays are 512×512. The duplicate all/ folder is intentionally not included to reduce release size.
Memory-SAM
This dataset accompanies Memory-SAM, a human-prompt-free retrieval-to-prompt framework for automatic tongue segmentation with SAM2. The included memory-sm_tongue/ folder provides the 20-item SM-Tongue memory bank used by the released default setting.
Privacy
Only anonymized numeric filenames are included. Original filenames are not included in this public release.
