Chart Classification Model

This project implements a deep learning model to classify whether an image contains a chart or not. This can be useful for filtering out non-chart images from a large set of images, and focusing on the subtask of chart analysis, such as chart structure extraction, data extraction, etc.

Project Structure

chart_classifier/
├── config.py             # Configuration files
├── data/            # Dataset and data processing
├── models/          # Model architecture
├── trainer/         # Training logic
├── utils/           # Utility functions
├── train.py         # Training script
└── inference.py     # Inference script

Setup

1. Clone the repository:

git clone https://github.com/beholder91/chart_classifier.git
cd chart-classifier

2. Install dependencies:

pip install -r requirements.txt

Training

To train the model, prepare your data first:

• Place chart images in data/dataset/charts/
• Place non-chart images in data/dataset/non-charts/

then run:

python train.py

Inference

To run inference on new images:

python inference.py

Model Architecture

• Base model: EfficientNet-B0
• Input size: 224x224
• Output: Binary classification (chart/non-chart)

Requirements

See requirements.txt for detailed dependencies.

License

MIT license