Image Classification with TensorFlow on macOS GPU

This project demonstrates how to build and train an image classifier using TensorFlow with GPU acceleration on macOS, leveraging Apple's Metal framework.

Features

Utilizes TensorFlow with Metal GPU acceleration for macOS
Implements a simple Artificial Neural Network (ANN) for image classification
Supports multiple classes: buildings, forest, glacier, mountain, sea, and street
Optimized for performance on Apple Silicon (M1/M2/M3) GPUs

Requirements

macOS 13 (Ventura) or later
Python 3.9+
TensorFlow 2.x
Apple Silicon Mac (M1/M2/M3) for GPU acceleration

Installation

Clone the repository:

git clone https://github.com/adityaamehra/Image-Classifier.git
cd Image-Classifier

Create and activate a virtual environment:

python3 -m venv venv
source venv/bin/activate

Install the required packages:

pip install tensorflow-macos tensorflow-metal

Usage

Prepare your dataset:
- Organize images into train and test folders
- Each class should have its own subfolder
Update the data_dir variable in the script to point to your dataset location
Run the training script:
```
python train.py
```
Monitor the training progress and GPU utilization using Activity Monitor

Model Architecture

The model uses a simple ANN with the following layers:

Flatten layer (input: 150x150x3)
Dense layer (512 units, ReLU activation)
Dense layer (256 units, ReLU activation)
Output layer (6 units, Softmax activation)

Optimizations

Uses tf.data.AUTOTUNE for optimized data pipeline
Implements mixed precision training for improved performance
Leverages Metal GPU acceleration for faster training

Troubleshooting

If you encounter issues with GPU detection:

Ensure you're running macOS 13 (Ventura) or later

Reinstall TensorFlow dependencies:

pip install --upgrade tensorflow-macos tensorflow-metal

Verify Xcode command-line tools are installed:
```
xcode-select --install
```

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.