OpenVid-1M-mapping / README.md
phil329's picture
Update README.md
48add2e verified
metadata
license: cc-by-4.0
task_categories:
  - text-to-video
language:
  - en
tags:
  - text-to-video
  - Video Generative Model Training
  - Text-to-Video Diffusion Model Training
  - prompts
pretty_name: OpenVid-1M-mapping
size_categories:
  - 1M<n<10M

Summary

This is the extent dataset proposed in the paper "OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation". OpenVid-1M is a high-quality text-to-video dataset designed for research institutions to enhance video quality, featuring high aesthetics, clarity, and resolution. It can be used for direct training or as a quality tuning complement to other video datasets.

New Feature: Video-ZIP mapping files now available for efficient video lookup (see Dataset Structure and Extended Usage).

Dataset Structure

DATA_PATH
└── video_mappings (NEW!)
    β”œβ”€β”€ OpenVid_part0.csv
    β”œβ”€β”€ OpenVid_part1.csv
    β”œβ”€β”€ ...
    └── OpenVid_part185.csv
β”œβ”€β”€ README.md
└── ...

Key Updates

  1. Video-ZIP Mapping
    We provide 186 mapping files (OpenVid_part*.csv) containing:

    zip_file,video,video_path
    OpenVid_part0.zip,celebv_--Jiv5iYqT8_0.mp4,OpenVid_part1/celebv_--Jiv5iYqT8_0.mp4
    OpenVid_part0.zip,celebv_--Jiv5iYqT8_2.mp4,OpenVid_part1/celebv_--Jiv5iYqT8_2.mp4
    OpenVid_part0.zip,celebv_--QCZKgJt6o_0.mp4,OpenVid_part1/celebv_--QCZKgJt6o_0.mp4
    OpenVid_part0.zip,celebv_--oCWVOBuvA_0.mp4,OpenVid_part1/celebv_--oCWVOBuvA_0.mp4
    ...
    
    • zip_file: Source ZIP archive name
    • video: Video filename
    • video_path: Relative path within the ZIP file
  2. Mapping Features

    • Enables direct video lookup without full ZIP extraction
    • Contains cross-references between ZIP files and video paths
    • Split into multiple files for parallel processing

Extended Usage

1. Video Mapping Access

import pandas as pd

# Load all mapping files
mappings = [pd.read_csv(f"video_mappings/OpenVid_part{i}.csv") 
           for i in range(186)]
full_mapping = pd.concat(mappings)

# Find ZIP location for specific video
video_name = "celebv_--Jiv5iYqT8_0.mp4"
result = full_mapping[full_mapping.video == video_name]
print(f"Video {video_name} is in {result.zip_file.values[0]} at path {result.video_path.values[0]}")

2. Advanced ZIP Handling

Extraction Specific Video from Zip_file

# Extract specific video from ZIP without full decompression
unzip OpenVid_part0.zip "OpenVid_part1/celebv_--Jiv5iYqT8_0.mp4" -d target_folder
# Unzip files (-j flattens directory structure)

Citation and Acknowledgments

This mapping dataset was developed to enhance OpenVid-1M accessibility. Special thanks to the original dataset contributors.

@article{nan2024openvid,
  title={OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation},
  author={Nan, Kepan and Xie, Rui and Zhou, Penghao and Fan, Tiehan and Yang, Zhenheng and Chen, Zhijie and Li, Xiang and Yang, Jian and Tai, Ying},
  journal={arXiv preprint arXiv:2407.02371},
  year={2024}
}