describe-dataset 0.1.2

Add your description here

DescribeDataset

A Python tool for generating human-readable descriptions of dataset structures. This tool helps you understand the organization and content of your datasets by providing a clear, textual representation that can be used with LLMs or for documentation purposes.

Features

• 📁 Recursive folder structure exploration
• 📊 Support for multiple file formats:
  • Text files (.txt)
  • JSON files (.json)
  • YAML files (.yaml, .yml)
  • CSV files (.csv)
  • Images (.jpg, .jpeg, .png, .gif, .bmp, .webp)
• 🖼️ Image resolution information
• 📏 Smart list clipping for large directories
• 📝 Total item count for folders
• 🛡️ Error handling for corrupted files

Installation

pip install describe-dataset

Usage

Command Line

python -m describe_dataset /path/to/your/dataset

Python API

from describe_dataset import describe_dataset
from pathlib import Path

# Get dataset description
description = describe_dataset(Path("/path/to/your/dataset"))
print(description)

Example Output

Top level of the dataset contains:
    folder images
    folder annotations
    file config.yaml

Content of images folder:
Total items in folder: 1000 (showing first 3)
- image1.jpg (file)
  image of resolution 1920x1080
- image2.jpg (file)
  image of resolution 1280x720
- image3.jpg (file)
  image of resolution 800x600

Content of annotations folder:
- train.json (file)
{
    "version": "1.0",
    "images": [...],
    "annotations": [...]
}

Description of config.yaml file:
name: my-dataset
version: 1.0
classes:
  - car
  - pedestrian
  - bicycle

Configuration

The tool has several configurable parameters:

• DIFFERENT_FILES_LIMIT = 10: Maximum number of different files to show in a folder
• MIN_LIST_LENGTH = 3: Minimum number of items to show when clipping large lists
• LIST_SIZE_THRESHOLD_FOR_MIN_LENGTH = 20: Threshold for when to start clipping lists

Contributing

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

License

This project is licensed under the MIT License - see the LICENSE file for details.