Traffic Sign Classification using TensorFlow

Introduction

This project implements multi-class classification of traffic signs using tensorflow/keras. The code is easy and understandable for anyone learning deep learning and neural networks. the goal is to show how a real world application is created. Traffic sign detection is must-have component for self driving cars, smart transportation etc.

Features

• Custom preprocessing functions using opencv

• Automated data loading with class wise organisation

• Visualisation of samples

• CNN built with tensorflow/keras

• Generation loss and accuracy curves

• Generationg classification report

• Prediction on test data with confidence score

  

Libraries Used

• TensorFlow/Keras
• OpenCV
• Numpy
• Scikit-Learn
• Matplotlib
• Pandas

Important

• Any dataset can be used (such as GTSRB or your own custom dataset). The data folder should be like:

  • traffic_Data
    • DATA
      • 0
      • 1
      • ...
    • TEST
      • img 000_0001
      • img 000_0002
      • ...
  • labels.csv

• For preprocessing, ImageDataGenerator should ideally be used. OpenCV functions are used only for custom augmentation and manual preprocessing steps that show how each transformation works internally.