DSCT

data.zip contains 95 images with various crack numbers and shapes (e.g., single crack, multi-crack, strip crack, mesh crack). The image size is 300×300 pixels after re-size transformation. Two Chinese medicine practitioners collaboratively divided95 images into 42 severe cracked tongues (labeled 1) and 53 mild cracked tongues (labeled 0). We call these image labels the gold standard in this study. The image size is 300×300 pixels after re-size transformation. code.zip involves crack extraction, feature extraction and severe cracked tongue recognition with different features and different classifiers.