Traffic Sign Classification Dataset
Dataset of road sign images (58 classes), designed to train and test CNN classification models with CSV labeling.
Approximately 6,165 JPG/PNG images divided into 58 classes, + labels.csv, ~2,000 test files
CC0: Public Domain
Description
The Traffic Sign Classification Dataset includes approximately 6,165 images of road signs divided into 58 categories. A labels.csv file provides the correspondence between classID and description. Approximately 2,000 images are reserved for the test set. The dataset is compatible with classic CNN models.
What is this dataset for?
- Train panel image classification models for assisted driving
- Evaluate neural networks for real-time recognition (e.g. embedded systems)
- Serve as an educational base for computer vision projects
Can it be enriched or improved?
You can enrich this dataset by adding location annotations (bounding boxes), augmenting images by rotation/bloom, or supplement with data from other signal bases for greater robustness.
🔎 In summary
🧠 Recommended for
- AI vision students
- Autonomous driving researchers
- Deep learning trainers
🔧 Compatible tools
- TensorFlow
- PyTorch
- Keras
- OpenCV
- LabelImg
💡 Tip
Use test images to validate performance in real conditions and test robustness to changes in angle and brightness.
Frequently Asked Questions
Does this dataset contain panel location annotations?
No, it is only a CSV file with labels per image, without bounding boxes.
How many panel classes are represented?
The dataset includes approximately 58 different classes of road signs.
Can it be used for commercial purposes?
Yes, the CC0 license allows free use, including commercial use.
