Welcome to the 2025-2026 spring term iteration of CS156b! As with last year, we'll be working with a dataset of radiograph images. Radiography is the most commonly used method of chest imaging, and has been a staple tool for diagnosis and screening of diseases. Automated anomaly detection in radiograph images would greatly assist medical professionals with correctly diagnosing these anomalies.
Public phase: Submit to receive scores on the public test set. Your score is visible on the leaderboard and you can compare with other participants.
Private phase: Submit to be evaluated on the private (holdout) test set. Your private score is not shown to you or other participants until the challenge ends. The admin can see the private leaderboard.
File format: Submit a .csv file.
Required columns (in any order): Id, No Finding, Enlarged Cardiomediastinum, Cardiomegaly, Lung Opacity, Pneumonia, Pleural Effusion, Pleural Other, Fracture, Support Devices.
Scoring: For each category, the Mean Squared Error (MSE) is computed and scaled by the variance to produce a normalized score. Blank labels in the ground truth are excluded from scoring. Lower scores are better.
Honor Code: No member of the community shall take unfair advantage of any other member of the community.
Resources: You may use online resources (papers, articles, code) as long as they do not directly pertain to this specific challenge.
| No | username | Cardiomegaly | Enlarged Cardiomediastinum | Fracture | Lung Opacity | No Finding | Pleural Effusion | Pleural Other | Pneumonia | Support Devices | Average | total_submission | last_sub |
|---|