A simple Python-based image classification application built as part of a machine learning and computer vision learning exercise.

The project uses a pre-trained ResNet-50 model from PyTorch/TorchVision to classify images against the ImageNet dataset and display the top-5 predicted classes.

Features

• Load a pre-trained ResNet-50 model
• Use ImageNet-trained weights
• Select images using a native macOS file picker
• Generate random placeholder images for testing
• Apply standard ImageNet preprocessing and normalization
• Run inference using PyTorch
• Display top-5 ImageNet predictions with probabilities
• Optional Core ML model export for macOS/iOS experimentation

Image ↓ Load Image ↓ Apply Transform ↓ Tensor [1, 3, 224, 224] ↓ ResNet-50 ↓ Softmax ↓ Top-K Selection ↓ Human-Readable Labels

Requirements

• Python 3.12+
• PyTorch
• TorchVision
• Pillow
• NumPy

Optional:

• coremltools (for Core ML export)

Create a virtual environment uv venv source .venv/bin/activate