showlab / UniVTG

[ICCV2023] UniVTG: Towards Unified Video-Language Temporal Grounding

Home Page:https://arxiv.org/abs/2307.16715

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Question about language-evaluation's version

APiaoG opened this issue · comments

Hello! Thank you for your outstanding work!
I encountered the following error while configuring the environment. Which version can I switch to?
image

@APiaoG Thanks for your interesting!

For this issue, please simply remove langchain in requirement list and try it again!

@APiaoG Thanks for your interesting!

For this issue, please simply remove langchain in requirement list and try it again!

Hello, thank you very much for your reply! I have removed the langchain but still encountered this issue, as shown in the following figure.
image

At the same time, I have discovered other issues.
image

As I have not encountered these issues or had any experience using them before, I am not sure if they are necessary or if changing the Python version and their versions will affect the results. Looking forward to your reply!

Hi @APiaoG ,
Thanks for mentioning this issues,
I have re-updated the requirement, and have removed the langcodes and nltk. can you re-try it and reach out to me?

Thank you again for your reply! The newly uploadedrequirement.txt is currently installed without any issues! I haven't downloaded the new data features yet, so I'm not sure if there will be any issues during training. Also, I would like to ask how I could quickstart the result reproduction work? May I also ask if it's enough for me to install the dataset according to the prompts in the file and run main_gradio.py directly? Thank you very much for your timely reply. You have helped me a lot! Thank you again for your reply!

Hi @APiaoG ,
for a quick start, main_gradio.py is the best way, you just need to download the checkpoint based on instructions and do not need to download any others, then you should able to perform temporal grounding on your mp4 video, this will be very easy.

but if you want to reproduce the paper result, you need to download the video / text features, and you might require some debugging for training since the current codes might not very clean.

I encourage you to try the main_gradio first and can download the video/text features for preparation.
I will continue to clean my codes at the same time.

close since have resolved the issues.