InternVideo-Next MobileClip-B

[xuridi01]

Hello,

First of all, I would like to thank you for your excellent work on InternVideo-Next. I found the paper very insightful.

In the paper, you mention the model’s zero-shot video–text retrieval capability using MobileCLIP-B. I was wondering whether you plan to publish or release the text encoder that was used for this setup.

Thank you very much for your time and consideration.

[getmoments-com]

That would be great!

[toretak]

I would really appreciate it too.

[Revliter]

Hello,
Thank you for your interest and your kind words about our work.
Unfortunately, due to a recent personal job transition, I was only able to retain the critical backbone weights. I did not save the specific checkpoints for the probing tests on the frozen model representations.
However, we plan to release both the training and probing code upon the official publication of the paper, so please stay tuned for that. In the meantime, the retrieval probing test with MobileCLIP-base largely follows the implementation here: https://github.com/OpenGVLab/InternVideo/tree/main/InternVideo2/multi_modality. You can refer to this if you need to reproduce the setup immediately.