search

pixeli99 / TrackDiffusion

Official PyTorch implementation of TrackDiffusion (https://arxiv.org/abs/2312.00651)
63 stars 4 forks source link

Inquiry about the release of web_demo.py #2

Closed zhtjtcz closed 7 months ago

zhtjtcz commented 7 months ago

I'd like to ask when the web_demo.py will be released? And using OpenAI key to run the demo is a bit of a tall order for many people, but are there any plans for a simplified version?

pixeli99 commented 7 months ago

Yes, I am trying to solve the user interaction problem with a lower-cost solution, because our approach requires specifying the trajectory of the entire segment for the target, which is indeed not very convenient for interaction. In the released version, I will use the api of gemini1.0-pro instead of asking everyone to enter the OpenAI key.

This week, web_demo will be released along with the weights.

KaiChen1998 commented 7 months ago

@zhtjtcz Thank you for your interest in our work! The training code and the pre-trained checkpoints of both text- and image-conditional TrackDiffusion have been released. The web demo is still on the way. Stay tuned for the latest updates :)!

qiuhaining commented 7 months ago

@zhtjtcz Thank you for your interest in our work! The training code and the pre-trained checkpoints of both text- and image-conditional TrackDiffusion have been released. The web demo is still on the way. Stay tuned for the latest updates :)!

where is the "script.sh"? I find other bugs: 1、not find "print_log" https://github.com/pixeli99/TrackDiffusion/blob/main/dataset/youtube_loader.py#L221 2、not find I2V codes 3、not find train.json in youtobevis 2019 datasets from https://opendatalab.com/YouTubeVIS2019/download