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

[Miayui]

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]

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

[Miayui]

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

[wenyuqing]

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

[Miayui]

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 [email protected], please cheak it~

[wenyuqing]

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

[Miayui]

Maybe the mail is in the trash?

[wenyuqing]

Hi, I'm very sorry I've checked all the mail but cannot find your email. Or maybe can you attach some images here?