metadata
license: cc-by-4.0
Caltech Football Numbers (CaltechFN) Cleaned
This version has been meticulously cleaned to improve data quality for model training. A Vision Large Language Model (google/gemma-3-4b-it) was used to verify the label of each image. Images that were corrupted, unclear, or had incorrect labels were removed.
The final filtered dataset contains:
- Training set: 100,442 images
- Test set: 25,077 images