akshayb80 / multi-class-vehicle-classification

Source Code for Multiclass Vehicle Detection & Classification

Multi-Class-Vehicle-Classification

A system that can detect and classify vehicles using deep learning.

Table of Contents

• About the Project
  • Tech Stack
  • File Structure
• Getting Started
  • Installation
• Results and Demo
• Future Work
• Troubleshooting
• Contributors
• Acknowledgements and Resources

About The Project

Our Project aims to analyse a large dataset of images containing various vehicle categories. We have built a Convolutional Neural Network utilizing LeNet Architecture to detect and classify vehicles from mulitple angles. The architecture consists of two sets of convolutional and average pooling layers, followed by a flattening convoutional layer, then two fully-connected layers finally use a softmax classifier.

Refer this documentation

Tech Stack

• OpenCV
• Jupyter Notebook
• Pycharm

File Structure

.
├── docs                    # Documentation files
│   ├── report.pdf          # Project report
│   └── results             # Video feed of the Working Model
├── src
    ├── hell.model          # CNN Model
│   ├── main.py             # Main File
│   └── train_network.py    # Training Network
├── ...
├── test                    # Test files
│   ├── test_network.py     # Testing
├── ...
├── LICENSE
├── README.md 
├── Setup.md                # Installation
└── todo.md                 # Future Developments

Getting Started

Prerequisites

• Anaconda Environment

  You can visit the Anaconda Website for the installation packages.

• Tensorflow-GPU version 2.1.0 (GPU version is recommended for faster performance)

  Tensorflow installation in Conda Environment

  Command for One Step installation (If the system has NVIDIA GPU):

conda create --name tensor_gpu tensorflow-gpu anaconda

• OpenCV version 4.3.0

conda install -c conda-forge opencv

Installation

1. Clone the repo

git clone https://github.com/akshayb80/Multi-Class-Vehicle-Classification.git

Results and Demo

A video demonstrating our working model
Working Model Video

Future Work

• Integrate this project with the License Plate Rocognition System

Troubleshooting

• Mulitple epochs required to get the best accuracy
• Ensure there is no Tensorflow compatibilty issues with the GPU before training

Contributors

• Akshay Bakshi
• Purvank Bhiwgade
• Tushar Bauskar
• Mohak Chandani

Acknowledgements and Resources

• SRA VJTI Eklavya 2020
• Refered OpenCV Tutorials for Image Processing
• Refered MIT Deep Learning for bulding Neural Network Model
• Refered Pyimagesearch for additional OpenCV materials ...