wuminye / nhr Goto Github PK

Hi,
I wanna train this project on my own dataset, so I wonder how the camera extrinsics are arranged in the file CamPose.inf?
I read the file tools/render.ipynb, where the cosine of the Angle of rotation about the Z-axis, X-axis, Y-axis are arranged in order in the direction of the column. Is this the correct order of the input camera extrinsics to the training process?
I wonder whether the camera intrinsics in file Intrinsic.inf, the number in matrix [0,1] is caused by camera distortion?
And what is the parameter whose name is INPUT.NEAR_FAR_SIZE in the file config.yml?

Since my dataset needs to be adjusted in accordance with the format required by the training input, I would like to consult you about these matters. Thank you so much~

Hi Mr, Wu

The point cloud files and some frames of images are missing in the released dataset.

Hi,did you have any data.
My own data set still doesn't fit the requirements.
I would like to learn about these during project implementation recurrence

你好！我根据GitHub的readme先运行了train_net.py，将模型训练至第30epoch，紧接着我运行了以下步骤

Network Fine-tuning
run cd tools && python finetune.py <gpu id> <path to checkpoint> <the number of resuming epoch>

但是我并没有发现这一步有输出，并且这一步很快就结束了，我的终端上只显示了logging的打印内容

0 / 200
50 / 200
100 / 200
150 / 200
load 56 Ts, 56 Ks, 200 frame, 33713011 vertices
dataset initialed.
2021-03-18 16:48:28,755 rendering_model.train INFO: Load 1 datasets.
2021-03-18 16:48:30,722 rendering_model.train INFO: Load pretrain model nr_model_30.pth.

然后程序执行就结束了，并且config中的OUTPUT_DIR里并没有新的文件。请问Network Fine-tuning这一步是有什么输出吗？
我觉得这里可能存在一些问题，但是我并不知道问题出在哪里。

Hi,

I want to know how the parameter MUL_POINTNET works?
I found the training process will be Nan when I train one dataset with MUL_POINTNET=True, but I found in code it is the same as MUL_POINTNET=False when there is only one dataset.

Looking forward to your reply. Thanks!

Best regards,
Leo

Could you release the codes about "Shape-from--Silhouettes" and "Hole patching"?

Thanks for your excellent!

Can you provide codes for the quantitative evaluation?

When I ran train_net.py, I got an error

RuntimeError: expected scalar type Float but found Half (data_ptr at /home/cy/anaconda3/envs/render/lib/python3.6/s ite-packages/torch/include/ATen/core/TensorMethods.h:6321)
frame #0: c10::Error::Error(c10::SourceLocation, std::string const&) + 0x47 (0x7f32dc94a627 in /home/cy/anaconda3/envs/NHR /lib/python3.6/site-packages/torch/lib/libc10.so)
frame #1: float* at::Tensor::data_ptr() const + 0xa60 (0x7f32bbb33480 in /home/cy/anaconda3/envs/NHR/lib/python3.6/ site-packages/pointnet2-0.0.0-py3.6-linux-x86_64.egg/pointnet2_cuda.cpython-36m-x86_64-linux-gnu.so)
frame #2: group_points_wrapper_fast(int, int, int, int, int, at::Tensor, at::Tensor, at::Tensor) + 0x27 (0x7f32bbb34427 in /home/cy/anaconda3/envs/NHR/lib/python3.6/site-packages/pointnet2-0.0.0-py3.6-linux-x86_64.egg/pointnet2_cuda.cpython-36m -x86_64-linux-gnu.so)
frame #3: + 0x14947 (0x7f32bbb2c947 in /home/cy/anaconda3/envs/NHR/lib/python3.6/site-packages/pointnet 2-0.0.0-py3.6-linux-x86_64.egg/pointnet2_cuda.cpython-36m-x86_64-linux-gnu.so)
frame #4: + 0x11a06 (0x7f32bbb29a06 in /home/cy/anaconda3/envs/NHR/lib/python3.6/site-packages/pointnet 2-0.0.0-py3.6-linux-x86_64.egg/pointnet2_cuda.cpython-36m-x86_64-linux-gnu.so)

frame #11: THPFunction_apply(_object*, _object*) + 0xa1f (0x7f330e1f6e3f in /home/cy/anaconda3/envs/NHR/lib/python3.6/site -packages/torch/lib/libtorch_python.so)

then I found that the error occurred on the line 40 of pointnet2_modules.py, which is
new_features = self.mlps[i](new_features).
The type of ‘new_features’ as input variable is 'torch.float32', but the type of output of 'self.mlps' is 'torch.float16'.

/NHR/modeling/UNet/unet_parts.py in
4 import torch.nn as nn
5 import torch.nn.functional as F
----> 6 import models_lpf
7
8

ModuleNotFoundError: No module named 'models_lpf'

ENVs:

Ubuntu16
CUDA10