GTSRB Traffic Vision with LeNet-5 (modified)

An image classifier using Convolutional Neural Network (CNN) to predict German traffic signs.

source

Overview dataset

German Traffic Sign Recognition Benchmark (GTSRB) is a large multi-category classification benchmark. The first was used in a competition at IJCNN 2011. The images are classified into 43 classes, consisting of 26640 training labeled images and 12630 testing set.

Models

There are few trained models with different configuration and CNN architecture

• GTSRBModelV0
• GTSRBModelV1 (work in progress)

Sample Predictions

model: GTSRBModelV0