wenyuqing / panacea

[CVPR2024] Official Repository of Paper "Panacea: Panoramic and Controllable Video Generation for Autonomous Driving"
https://panacea-ad.github.io/
Apache License 2.0
197 stars 12 forks source link

a issue about background like traffic cone or road line isn't constrained by the BEV layout #30

Open Miayui opened 4 weeks ago

Miayui commented 4 weeks ago

hello, thank you for your excellent work! I had a problem when I was trying to run inference.py. Though the model successfully output the video, but in this video, only the instance like car was constrained by BEV layout, the background like traffic cone and road line wasn't constrained by BEV layout, meanwhile the ego vehicle remained stationary in video. I have already check the input 19 channels (10 for depth, 3 for bounding boxes, 3 for road maps, and 3 for camera pose embeddings) for control net and didn't find any mistake. Could you help me out with this? thank you!

wenyuqing commented 3 weeks ago

Could you give me some examples for that so I can check the possible problem?

Miayui commented 3 weeks ago

Thank you so much for your reply! I've already send some inference results to your email this morning from edith.huang1999@formail.com. The index for results in "nuscenes2d_temporal_infos_train.pkl" are [1827:1832] and [1828:1833].

wenyuqing commented 2 weeks ago

Hi, sorry for the late reply, it seems that I didn't receive your pictures, could you please send them one more time?

Miayui commented 2 weeks ago

hi, thank you for your reply, I can't send the pictures to your github mail cause they are too big, so I send to your school e-mail wenyuqing@mail.ustc.edu.cn, please cheak it~

wenyuqing commented 2 weeks ago

Hi, sorry, maybe there is something wrong? I cannot find your email.

Miayui commented 2 weeks ago

Maybe the mail is in the trash?