search

OpenGVLab / Ask-Anything

[CVPR2024 Highlight][VideoChatGPT] ChatGPT with video understanding! And many more supported LMs such as miniGPT4, StableLM, and MOSS.
https://vchat.opengvlab.com/
MIT License
3.05k stars 251 forks source link

Release of GPT scripts to generate the QA data for MVBench #229

Closed BennoKrojer closed 3 weeks ago

BennoKrojer commented 2 months ago

Hello,

Thanks for releasing MVBench and all the other great things! I am studying video understanding benchmarks in my research and was wondering if you could release the scripts/prompts for QA generation with GPT, e.g. for CLEVERER or Action Antonym?

Your help would be much appreciated! Even just the raw scripts+prompts I could work with.

Best, Benno

Andy1621 commented 2 months ago

Thanks for your great interest. Acutally, most of the QAs are generated by template. For example, the Action Antonym is builted on PAXION. We design the options and simply use GPT to generate simple questions.

One special example is Unexpected Action, I used ChatGPT to generate multiple-choice options from simple QA. The prompt is modified from FunQA.

yinanhe commented 3 weeks ago

Hi, we will close this issue. Feel free to contact us if you have other questions.

BennoKrojer commented 1 week ago

Hi,

thank you for your reply! It is still not clear to me how exactly Egocentric Navigation for example was created. How did you know when exactly to stop the video such that the next instruction is not shown in the video? Releasing these scripts would be quite helpful for me and maybe others! Thank you, Benno