This group project explores basic machine learning models (MLP, RF and SVM) in classifying traffic signs. The dataset used for this project is GTRSB (German Traffic Sign Recognition Benchmark).
The model with the best performance is MLP(300) on HOG features on non-processed image with accuracy of 93% and F1 score of 92%
- Includes train and test images in the dataset under .npy format. Images are separated into processed image and original images
- Due to file size restriction, the data for this project can be accessed through drive
Includes Jupyter notebook files of our 3 models (MLP, RF and SVM) trained on HoG features, SIFT+BoW features and on the images themselves
Includes the model with the best performance
Run python demo.py
to start the demonstration program. Upload a traffic sign and let the program predict the label
- Ngô Phương Nhi
- Lâm Văn Sang Em
- Đặng Chí Trung