5,964 training and 1,003 validation images of gear digits (0-9) extracted from racing car onboard video feeds. The dataset, created by tobil, is designed for real-time gear indicator reading and is supplemented with MNIST digits for robustness. It was last updated on March 31, 2026.
Use Cases
- Train a real-time gear number detector based on images from racing onboard cameras.
- Benchmark digit recognition models on a domain-specific application mentioned in the description.
- Develop data augmentation techniques using the mix of racing-original and MNIST-supplemented images.
- Fine-tune pre-trained models for automotive telemetry overlay understanding.
Strengths
- Contains 6,967 total labeled images across 10 classes.
- Images are a consistent 32x32 pixels in grayscale PNG format.
- Includes a source column to distinguish racing-original data from supplemented MNIST digits.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
Provenance
- Source
- huggingface user tobil
- Collection Method
- Likely extracted from racing onboard video telemetry overlays and supplemented with MNIST digits.
- Time Range
- null
- Freshness
- Last updated 2026-03-31 16:48:07; freshness should be verified.
- Geography
- null