Zheng Gong

Due to KITTI Semantic datasets' dimension is 4D(xyz+brightness),I tried to imitate a 4d version of chamfer,can you help me to inspect whether my code is correct?

1

Hi,@ThibaultGROUEIX Thanks for your work!My 4D version is as below by imitation. ```cuda #include #include #include #include #include __global__ void NmDistanceKernel(int b,int n,const float * xyz,int m,const float * xyz2,float...

Due to the implementation of Deepfool and C&W is for classifier which just have a 1xK dimension predict vector，but for segmentation model，the predict is a mask whose dimension is HxWxK.

How bayesian network be used in quantization?

1

@peteriz @jpablomch @ranganathkrishnan we can get the $\mu$ and $\sigma$ of each weight ,so how it can be used to quantization which mapping the $w_{float32 }$ into $w_{int8}$

hi, @bermanmaxim jaccard_loss = 1 - IOU why do the jaccard[1:] - jaccard[0:-1]? ```python def lovasz_grad(gt_sorted): """ Computes gradient of the Lovasz extension w.r.t sorted errors See Alg. 1 in...

when load pretrained model of VoVNet99, missing key in state_dict

1

hi, thanks for your work,but when I want to load the pretrained model, the key is not as the same as the model I build. I extract the model building...

Thanks for answering in the [https://github.com/Soumyabrata/CloudSegNet/issues/2#issue-1207608271](https://github.com/Soumyabrata/CloudSegNet/issues/2#issue-1207608271),but the logic of the code show that the list to be shuffled is the all augmented data list,which may lead to use the variants...

Hi, @percyliang , thanks for your sharing! But when I try to regenerate instruction by script `generate_instruction.py`, the cost of time seems so long, even if I reduce the number...

具体error 

Hi, @longouyang @welinder @avital , thanks for your work in PRM. But I want to ask a question is that what's the training pipeline of PRM？ For ORM, we can...

The Input need to be fixed in network?

1

Hi,thanks for your @ShiQiu0419 Due to I saw the network have a parameter "num_points" in network to do Fathest Point Sampling,so why not use point.shape[0] as a dynamic n_points? Because...