Deephome

I have the same question. It seems that the code implementation only tracks the objects with 3D detection. The 2D-only detections are not used to create tracks.

@TouchDeeper 谢谢你的回复！当theta很小，近似确实是这样的，如果为了计算简单，那就说的通了。 但是，请你看下下面这个博客的1.3节，确实存在`纯虚四元数的指数映射` 参考：[四元数](https://blog.csdn.net/hzwwpgmwy/article/details/84846016)

@kentang-mit Thank you for your reply! I already replaced torchpack.distributed in `tools/test.py` by `torch.distributed`. Specifically, I comment ``` # dist.init() # torch.backends.cudnn.benchmark = True # torch.cuda.set_device(dist.local_rank()) ``` and add ```...

@yangxh11 @kentang-mit Thank you for your suggestion! I'll have a try!

@dmanegold Sure, could you provide more information about LVI-EXC, such as paper, feature, optimization method etc.

@sunjiahao1999 Hi, do you have any idea to solve this problem?

@manueldiaz96 Well done!

> > Hi @Deephome and @syn0419, have you ever tried to export models without `--without-nms`? > > Yes, I encountered this issue before. Even after updating the project and following...

@syn0419 I only check the weight of backbone.image_model.stem.conv, which is the first convolution of the network. The mean error of first conv's feature map is about 2e-08. I'm not sure...

@wondervictor Thank you for your update on inference_with_postprocessing() function, but it doesn't solve the problem. I compared the image feature maps of three sclaes (i.e. outputs of stage 3/4/5) between...