addy369 / bosch

German Traffic Sign Recognition Website

German Traffic Sign

An Artificial Intelligence tool that predicts Traffic signs based on various pre-trained models and allows user to manipulate datasets.

This repo contains:

• A React-Flask Based ML Web App

Our Web Application

Explainable AI

GradCam Technique to identify mislabelling hotspots

Using TSNE plots to visualize and evaluate model performance

Key Features

• Create a complex Dataset
• Train additional images on the fly
• View model performances across different metrics
• Visualize model performance
• Get suggestions to various shortcomings in model training
• An explainable AI-based solution to comprehend network failures

---

Prerequisites

1. Git.
2. Node & npm (version 12 or greater).
3. A fork of the repo.
4. Python3 environment to install flask

Directory Structure

The following is a high-level overview of relevant files and folders.

backend/
├── backend/
│   ├── template/
│   └── app.py

└── frontend/
    ├── public/
    │   ├── index.html
    │   └── ...
    ├── images/
    │   └── logo.png
    ├── src/
    │   ├── assets/
    │   │   ├── css
    │   │   └── fonts...
    │   ├── components/
    │   │   ├── Sidebar 
    │   │   └── Navbars
    │   └── views/
 
         ├── routes.js
         ├── package.json
       ├── local_vm.sh
       └── .gitignore
       

Installation

Clone

• Clone this repo to your local machine using https://github.com/anjalisoni3655/Bosch

Steps to run backend

In order to install all packages follow the steps below:

1. Download the static folder from this drive: https://drive.google.com/file/d/149fh2lq7fT35RQVP5rmTgUfcYPorE9kX/view
2. Put it in the backend/
3. Move to backend folder
4. For installing virtual environment - python3 -m pip install --user virtualenv
5. Create A Virtual env - python3 -m venv env
6. Activate virtual env - source env/bin/activate
7. pip3 install -r requirements.txt
8. flask run

Steps To Set Up Frontend

1. Move to frontend folder
2. npm install
3. npm start

The model will be served on http://127.0.0.1:5000/

---

License

This project is licensed under the Apache License, Version 2.0.