Inquiry regarding dataset preparation and preprocessing for reproduction

[hunblingbling]

Thank you for sharing your insightful research. I am currently working on reproducing your study and would like to request some clarification regarding the dataset preparation.

Specifically, I have the following questions:

Dataset Acquisition: What is the recommended procedure for downloading the OpenVid-1M dataset?

File Preparation: How should I generate the openvid-1m.parquet file required to execute the build_rag_database.py script? If there are specific preprocessing steps or conversion scripts needed to format the raw data into this parquet file, could you please provide guidance on that?

I look forward to your guidance. Thank you for your time and for contributing to the community.

Best regards,

[flateon]

Thanks for your question and interest in reproducing our work.
The recommended procedure is:

1. Download the preprocessed annotation file from our Hugging Face repository: https://huggingface.co/MCG-NJU/MotionRAG
2. Download and extract the video files from the OpenVid Hugging Face repository.
   The openvid-1k.parquet file is already provided in our HF repo and can be used directly with build_rag_database.py. At the moment, we do not provide a separate conversion script.

感谢您的提问和对我们工作的关注。
推荐的步骤是：

1. 首先从我们的 Hugging Face 仓库下载我们预处理后的数据集标注文件：https://huggingface.co/MCG-NJU/MotionRAG
2. 然后从 OpenVid 的 Hugging Face 仓库下载并解压视频文件。
   openvid-1m.parquet 文件已经放在我们的 HF 仓库中，可以直接用于 build_rag_database.py。目前我们暂时没有提供单独的转换脚本。