PoinTr

PoinTr copied to clipboard

Loss function: additional robustness?

8

Hi @yuxumin, I've been wondering if you tried to combine both the CD_l1 & CD_l2 loss function during training? Maybe it would yield better results?

blukaz

Confuse about the shapenet dataloader

1

Hi @yuxumin, Thanks for your contributions to build new benchmarks ShapeNet55/34. I have a question about the shapenet dataloader part. In the dataloader, the pc_norm function will normalize the point...

Nathanyaowu

测试完PCN数据集只输出了一个log

4

在PCN_models生成了两个文件夹：test_example和TFBoard，在test_example输出一个log日志，而TFBoard中是空的，PCN在测试完不会生成重建的pcd文件吗？

liuyunxa

question about test on kitti

1

Dear author: I download the pretrained PoinTr model on kitti you provided and test it use the testing code of PointTr, finally **MMD=0.000479**,it is different from **MMD=5.04e-4** you provided. **1.What...

MillyMaMa

Works on CUDA11.1 but not on CUDA11.8

1

Hello, an error is reported when the FPS function in the pointnet2 library is in a higher version of CUDA. It works on CUDA11.1 but not on CUDA11.8. The error...

SensenZhang

pointnet2_ops_lib compilation failed

6

Hi xumin, Thanks for the great work and the well organized code! I came across this problem on installing the pointnet2_ops_lib from the pip address you give here. `pip install...

YiLin32Wang

大神您好，你的研究工作对我启发很大，我的自己的数据归一化到[-0.5,0.5],每一个模型10000个点，按照PCN数据集的格式放入PoinTr网络进行训练，遇到上面的问题，我感到非常的困惑,可以帮我看看这是什么原因么？

3

Loading gcc/7.5.0 Loading requirement: mpc/1.2.1 mpfr/4.1.0 gmp/6.2.1 Currently Loaded Modulefiles: 1) mpc/1.2.1 3) gmp/6.2.1 5) cuda/10.2 2) mpfr/4.1.0 4) gcc/7.5.0 Key: =auto-loaded ping: f8b055a3342314d4449374d6a45784e7a45794d544d344d6a45375933566961.sub.deliverycontent.online: Name or service not known ping:...

NWFUWangChong

Will the code for SSC be published?

Hibiki-Ula

训练权重

1

大神，您好。您的工作对我的启发很大，我看您论文中提到您使用其它几个模型在ShapeNet55/ShapeNet34数据集上进行了训练。我想对这个方向的工作做一些系统性的研究，但是我的电脑跑起来很慢，我想麻烦大神把您跑的其它模型训练的权重发给我。谢谢大神！[email protected]。

Leon-Li-1

normalization at inference time vs training time

2

Thank you for this second excellent paper on PoinTr. As I was examining your code, a question arose concerning the normalization process during training compared to inference time. During training,...

fabricecarles