Vehicle Tires Health Condition Detector

This repository contains the code for detecting vehicle tire conditions using a deep learning model. The model classifies tire images into two categories: "perfect" and "defective" using transfer learning with MobileNet_v2.

Project Overview

The primary objective of this project is to learn and utilize weights and Biases's different logs and artifacts in keras and tensorflow . Also to enhance vehicle safety by providing a reliable method for identifying tire defects. Tire-related crashes are a significant concern, and timely detection of defects can prevent accidents and ensure road safety. Click here to try this project in huggingface space

Key Features

• Binary Classification: Classifies tire images into "perfect" and "defective".
• Transfer Learning: Utilizes MobileNet_v2 for efficient and accurate predictions.
• Real-Time Detection: Provides quick and reliable classification results.
• Web Interface: Includes a Gradio web interface for easy image upload and prediction.

Dataset

The dataset consists of images of vehicle tires categorized into "perfect" and "defective". The data is split into training, validation, and test sets. The dataset was obtained from Roboflow

Provided by a Roboflow user License: CC BY 4.0

Model Architecture

The model is built using MobileNet_v2 as the base model, with additional custom layers for the binary classification task.

Experiments Tracking with Weights and Biases

This project utilizes Weights and Biases to track experiments, log metrics, and visualize model performance.

Setup Weights and Biases

1. Create an account on Weights and Biases.
2. Install the wandb library:

   pip install wandb

3. Login to your Weights and Biases account:

   wandb login

4. Initialize Weights and Biases in your training script:

   import wandb
   wandb.init(project="Project name", entity="your_wandb_entity")
   
   # Example of logging metrics
   wandb.log({"accuracy": accuracy, "loss": loss})

Installation

To run this project, follow these steps:

1. Clone the repository:

   git clone https://github.com/Arsalanzabeeb786/vehicle-tire-condition-detection.git
   

2. Install the required dependencies:

   pip install -r requirements.txt

Report of weight and biases can be seen here

Contributing

Contributions are welcome! Please fork this repository and submit a pull request for any enhancements or bug fixes.

License

This project is licensed under the MIT License.