🛞 Tire Classification using Transfer Learning (ResNet50)

📖 Overview

This project predicts tire classes (e.g., damaged vs. normal) using Transfer Learning with a pretrained ResNet50 architecture from PyTorch. It demonstrates a modern deep learning pipeline including:

• 🧠 Pretrained ResNet50 backbone from ImageNet for feature extraction
• 🧩 Custom Fully Connected (FC) Classifier Head with Batch Normalization, ReLU, and Dropout
• ⚖️ Cross-Entropy Loss for binary classification
• 🚀 Adam optimizer for efficient training
• 🔒 Frozen ResNet backbone to leverage pre-learned visual features
• 🧰 Modular design — easily switch between saving only the FC head or the full model
• 🎨 Interactive Gradio Interface for real-time image classification visualization

---

🖼️ Application Screenshot

Below is a preview of the Gradio Interface used for real-time classification:

---

🧩 Libraries

• PyTorch – model, training, and inference
• torchvision – pretrained ResNet50 weights
• pandas, numpy – data handling
• matplotlib – loss visualization
• pickle – saving/loading normalization params and trained model
• Gradio — interactive web interface for real-time model demos

---

⚙️ Requirements

• Python 3.13+
• Recommended editor: VS Code

---

📦 Installation

• Clone the repository

git clone https://github.com/hurkanugur/ResNet-Tire-Classifier.git

• Navigate to the ResNet-Tire-Classifier directory

cd ResNet-Tire-Classifier

• Install dependencies

pip install -r requirements.txt

---

🔧 Setup Python Environment in VS Code

1. View → Command Palette → Python: Create Environment
2. Choose Venv and your Python version
3. Select requirements.txt to install dependencies
4. Click OK

---

🗂️ Download and Prepare Dataset

1. Download Dataset: 🔗 Click here to download
2. Unzip data.zip into the project directory
3. The data folder should be on the same hierarchy as src and model.

---

📂 Project Structure

assets/
└── app_screenshot.png                # Screenshot of the application

data/
├── defective                         # Defective tire images
└── good                              # Good tire images

src/
├── config.py                         # Paths, hyperparameters, split ratios
├── dataset.py                        # Data loading & preprocessing
├── device_manager.py                 # Selects and manages compute device
├── train.py                          # Training pipeline
├── inference.py                      # Inference pipeline
├── model.py                          # Neural network definition
└── visualize.py                      # Training/validation plots

main/
├── main_train.py                     # Entry point for training
└── main_inference.py                 # Entry point for inference

requirements.txt                      # Python dependencies

---

📂 Model Architecture

ResNet50 (Pretrained on ImageNet)
  ↓
[Feature Extractor]                   # All convolutional layers (frozen)
  ↓
Custom Classifier Head:
  → Linear(in_features, 1024)
  → BatchNorm1d(1024)
  → ReLU
  → Dropout(0.5)
  → Linear(1024, 2)
  → Softmax(Output)

---

📂 Train the Model

Navigate to the project directory:

cd ResNet-Tire-Classifier

Run the training script:

python -m main.main_train

or

python3 -m main.main_train

---

📂 Run Inference / Make Predictions

Navigate to the project directory:

cd ResNet-Tire-Classifier

Run the app:

python -m main.main_inference

or

python3 -m main.main_inference