A CNN based model constructed using tensorflow 2.0 and Keras that can detect german traffic signs.
·
Report a Bug
·
Request a Feature
This is one of my personal projects, used so I can learn more about a particular subject.
This is my take on a competition long ago closed, but that is still relevant as a beginners project of machine learning and multiclass classification.
My project was built using Tensorflow 2.0 with the built-in Keras framework. The architecture of my network is a simple CNN with multiple groups of two convolutional layers and one max pooling, I used intercalated dropout layers to dimish the overfitting of the model. It as a 97% accuracy in the test group.
As a necessity, I coded two scripts to create three groups from the main data set (training, validation, and test) their use is explained below.
You will need the GTSRB dataset and Tensorflow 2.0 to execute all of the scripts. If you only want the model simple copy the gtsrb_model.h5 file.
- Clone the repo
git clone https:://github.com/dario-santos/CNN-Machine-Learning-Model.git- Download the GTSRB dataset (link in the end)
- Run the prepare_data.py script
- Run the model.py to train the model It's recommended to use a GPU as it's fastest
- If you want to test your model in the test dataset you can do it by running the model_predict.py script.
If you only want the model it's the .h5 file.
See the open issues for a list of proposed features (and known issues).
Distributed under the GNU GPLv3 License. See LICENSE for more information.
Dário Santos - LinkedIn - [email protected]
Project Link: https://github.com/dario-santos/CNN-Machine-Learning-Model