ProgramerSalar/video-dataset

Video Dataset on Hugging Face

This repository hosts the video dataset, a widely used benchmark dataset for human action recognition in videos. The dataset has been processed and uploaded to the Hugging Face Hub for easy access, sharing, and integration into machine learning workflows.

---

---

Introduction

The dataset is a large-scale video dataset designed for action recognition tasks. It contains 13,320 video clips across 101 action categories, making it one of the most comprehensive datasets for video-based machine learning research. The dataset is widely used in academia and industry for benchmarking video classification models.

This repository provides a Hugging Face-compatible version of the dataset, enabling seamless integration with the Hugging Face ecosystem. By hosting the dataset on the Hugging Face Hub, we aim to simplify dataset access, sharing, and reproducibility for researchers and developers.

---

Dataset Overview

• Name:
• Number of Classes: 101
• Number of Videos: 13,320
• Categories: Human actions such as sports, musical instruments, and human-object interactions.
• Format: The dataset is stored in an image folder format, where each class has its own subdirectory containing frames extracted from videos.

The dataset is ideal for tasks such as:

• Action recognition
• Video classification
• Transfer learning
• Benchmarking deep learning models

---

Dataset Structure

The dataset is organized in an image folder format, which is compatible with the Hugging Face datasets library. The directory structure is as follows:

/data
├── class1
│   ├── video1_frame1.jpg
│   ├── video1_frame2.jpg
│   └── ...
├── class2
│   ├── video2_frame1.jpg
│   ├── video2_frame2.jpg
│   └── ...
└── ...

Each subdirectory represents a class, and the images within the subdirectory are frames extracted from videos belonging to that class.

---

Usage

Loading the Dataset

To load the dataset using the Hugging Face datasets library, use the following code:

from datasets import load_dataset

dataset = load_dataset("ProgramerSalar/video_dataset")

Accessing the Dataset

Once loaded, you can access the dataset's splits and features:

print(dataset)

Pushing Updates to the Dataset

If you make modifications to the dataset and want to push the changes to the Hugging Face Hub, use the following code:

dataset.push_to_hub("ProgramerSalar/video_dataset")

---

Features

• Easy Access: Load the dataset directly from the Hugging Face Hub without manual downloads or preprocessing.
• Hugging Face Integration: Compatible with the Hugging Face datasets library for seamless integration into machine learning pipelines.
• Customizable: Modify the dataset structure or preprocessing steps to suit your specific needs.
• Reproducibility: Ensure consistent dataset handling across different projects and teams.

---

Preprocessing

The dataset has been preprocessed to extract frames from videos and organize them into an image folder format. This format is widely used for video classification tasks and is compatible with most deep learning frameworks.

If you wish to preprocess the dataset further, you can use tools like OpenCV or FFmpeg to extract frames, resize images, or apply data augmentation.

---

Hugging Face Integration

The Hugging Face Hub is a powerful platform for sharing datasets, models, and other machine learning assets. By hosting the dataset on the Hub, we provide the following benefits:

• Centralized Access: Access the dataset from anywhere with an internet connection.
• Collaboration: Share the dataset with collaborators or the community.
• Version Control: Maintain different versions of the dataset for reproducibility.

To push updates to the dataset, authenticate with your Hugging Face account and use the push_to_hub method provided by the datasets library.

---

Applications

The dataset is suitable for a wide range of applications, including:

• Action Recognition: Train models to classify human actions in videos.
• Transfer Learning: Use the dataset to pretrain models for video-based tasks.
• Benchmarking: Evaluate the performance of video classification models.
• Research: Explore new architectures and techniques for video understanding.

---

Contributing

Contributions to this project are welcome! If you have suggestions for improvements or new features, feel free to open an issue or submit a pull request. Please ensure your contributions adhere to the following guidelines:

• Follow the existing code style and structure.
• Provide clear and concise documentation for any new features.
• Test your changes thoroughly before submitting.

---

License

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

---

Acknowledgments

• ** Dataset Creators**: Thank you for creating and maintaining this invaluable resource for the machine learning community.
• Hugging Face: For providing the datasets library and the Hub platform.
• Open-Source Community: For contributing tools and resources that make machine learning accessible to everyone.

---

Contact

If you have any questions or need assistance, feel free to reach out:

• Author: ProgramerSalar
• Email: [[email protected]]
• GitHub: ProgramerSalar

---

Happy coding and exploring the dataset! 🚀