Yuqing Wen
Yuqing Wen
请问您这里是不是没有生成.mat文件的代码啊
> > > > > > > > > So, the result of BEVFormer-S in original paper is trained under `TemporalSelfAttention` while tested under degraded SelfAttention(with `video_test_mode=False` in config) ?...
Thanks for your quick reply! So how should I change this? Could you please provide more details if possible? I'm still a little confusing.
Hi，because the unzipped files from this zip archive may contain this absolute paths, please ensure that the relative paths for the dataset are correct.
If no further problem, I'll close this issue temporally.
Sorry for the late reply, the StreamPETR-s is obtained by closing the temporal propagation, which you may refer to the original StreamPETR paper for more information. I'm sorry that I...
Hello, theoretically it might not be sufficient. Maybe you should check the original implementation in StreamPETR for more specific details.
The current code use the GT image as conditional frame and generate the subsequent video frames for inference, so modifying the text prompt cannot modify the textual attributes well because...
Hi, the video frames generated are largely influenced by the image condition ( eg. the first/last frame), so the attributes controlling should be conducted on the generation of the first...
Could you give me some examples for that so I can check the possible problem?