Local dataset quality check
Detect train-test leakage in image datasets
The same image can appear across train, validation, and test splits, making results look better than they are.
What to check
Compare local image evidence and split mappings to find exact cross-split duplicates.
Why it matters
Leakage invalidates evaluation because the model has already seen test examples.
When Pro helps
Pro adds saved repair sessions for repeat review.
What your report can look like
Split leakage review
An illustrative example of the findings and repair guidance Labelyst produces for your dataset.
Status: Blocked
Score: 58/100
2 findings
Findings
- 2 identical images appear in train and test
- 1 split mapping needs review
Passed check
- Class labels were read from the selected export
Suggested next step: Move duplicate images out of the evaluation split.
Rules: standard profile · local check
Illustrative local result. It is not a certification.
Check it before training
No account or upload is required for a Free check. Analysis, repairs, and reports stay on your device.