Training CNN model for classifying traffic sign
The dataset using in this project is GTSRB - German Traffic Sign Recognition Benchmark from Kaggle.
Link: gtsrb-german-traffic-sign
The German Traffic Sign Benchmark is a multi-class, single-image classification challenge held at the International Joint Conference on Neural Networks (IJCNN) 2011. The dataset has the following properties:
- Single-image, multi-class classification problem
- 43 label classes
- More than 50,000 images in total
- Including train set, test set and meta data
- Large, lifelike database
2 models was built for testing the accuracy, with the best result (around 98% accuracy) came from the model with more hidden layers (and more neuron).
The model was exported to .h5 file for later use. The .h5 file can be downloaded via model_1_checkpoint.h5.