This project demonstrates the process of fine-tuning a pre-trained deep learning model (ResNet50) for multi-class image classification. The dataset is split into training, validation, and test sets, and the model is evaluated using metrics such as accuracy, precision, and AUC.

• Project Overview
• Dataset
• Model Architecture
• Training and Evaluation
• Results
• Installation
• Usage
• License

The goal of this project is to fine-tune a ResNet50 model pre-trained on ImageNet for a custom image classification task with 5 classes. Key features include:

• Data augmentation for robust training.
• Use of advanced metrics: Accuracy, Precision, Recall, F1-Score, and AUC.
• Evaluation on a well-defined test set.

The dataset used for this project is a custom dataset consisting of 5 classes of images. It is structured as follows:

The dataset is preprocessed using TensorFlow's image_dataset_from_directory utility to generate training and test batches.

• Base Model: ResNet50 pre-trained on ImageNet.
• Custom Layers:
  • Global Average Pooling
  • Dense Layer (512 units, ReLU activation)
  • Dropout Layer (50% dropout rate)
  • Output Layer (Softmax activation for multi-class classification)

1. Training:

   • Optimizer: Adam (Learning rate = 1e-5)
   • Loss: Categorical Crossentropy
   • Metrics: Accuracy

2. Evaluation:

   • Metrics: Accuracy, Precision, AUC
   • Additional insights: Classification Report

• Accuracy: 83.4% (Test Dataset, Batch Size = 16)
• Precision: 82.9%
• AUC: 0.81 (One-vs-All)

1. Clone the repository:

   git clone https://github.com/your-username/your-repo-name.git
   cd your-repo-name

2. Install required dependencies:

   pip install -r requirements.txt

3. Download the dataset and place it in the dataset/ directory following the structure mentioned above.

1. Run the script to fine-tune the model:

   python train_model.py

2. Evaluate the model:

   python evaluate_model.py

3. View the results and metrics in the terminal output.

This project is licensed under the MIT License. See the LICENSE file for more details.

• The ResNet50 model used in this project is pre-trained on ImageNet.
• TensorFlow and Keras were used for deep learning model development.
• Dataset sourced from [mention your source if applicable].