FastAI Experiments

Overview

Experiments using the FastAI framework for image classification.

The goal of this project was to:

• understand transfer learning
• experiment with data augmentation
• evaluate model performance on a custom dataset

Tech Stack

• Python 3.10
• fastai
• PyTorch

Dataset

Brief description (source, size, type).

Training Approach

• Pretrained ResNet
• Fine-tuning
• Data augmentation

Results

Accuracy: 0,87 Validation loss: 0,42

Reproducibility

conda env create -f environment.yml jupyter notebook