Tech-Meet-2021

Traffic Sign Recognition

We have worked on the high prep statement- German traffic sign recognition model along with a User Interface frame which itself allows user to augment test images and introduce new 5 classes.

This is a Deep Learning model developed using keras. It is used to recognise traffic signals. It was initially on the German Traffic Sign Recognition data set which was the baseline for this model. Later, 5 new classes were added to this model and the model was retrained to increase the difficulty of the data set. Then using an app called streamlit, the UI was created that enabled the user to play with the augmentations and visualise the results in terms of the accuracy and loss metrics. UI provided is very user friendly.

Setup

To setup the model in your local environment please go through the following steps:

1. Make sure you have Python 3.8 installed in your system. If not, install it from here and add its path to the environment variables.

To check if the python is correctly setup in the enviroment run python --version or python3 --version. It should print the installed version of python in your system.

2. Run the following code to install Streamlit (It is an open source python library that we have used to integrate the model with UI):
   pip install streamlit
3. We have used the following open source libraries. So, if you don't have them installed you need to install them by running pip install [library name]:

pandas
numpy
matplotlib
opencv_python
Pillow
tensorflow
keras
skimage

4. Clone this repository in your local system.
5. Download the saved models for different types of split into validation sets from here (Put all five directories in a folder saves).
6. Move this saves folder in the root directory of the cloned repository in your system.
7. Download the validation datasets from here: Test_final, Test_final_2.
8. Exctract these two directories Test_final and Test_final_2 in the root directory of the cloned repository in your system.
9. Open terminal and go to the location of the root directory of cloned repository in your system and run the following command:
   streamlit run bosch.py