yhw-yhw / pvamvsnet Goto Github PK

Hi, I cannot download from baiduyun. Can you put the files on google drive or github release, or can you send to my email? Thank you.

如题

作者您好，我在测试的时候使用1080ti和titanxp均显示显存不足，不知道是什么原因？测试使用的图片和模型都是从您提供的网盘上下载的，盼解答，谢谢。

Hi, @yhw-yhw

Thanks for your interesting work. The test result on Scan9 is incredible great.

But I was stucked when trying to test on my own dataset.

I tried to change the max_w and max_h , but it didn't work. And it seems that those params actually not be used in eval.py.

The shape of my input image is 16^x.

Does the eval.py only support for images(1200 x 1600) in dtu dataset? AND what should I do to test on other dataset?

By the way, I am confused, why you crop image in every pyramid? (in the orgin lvl, from 1200 x 1600 to 1184 x 1600).

Thanks.

you use volumegatelight here

which is defined as

but there is no sigmoid. Generally the attention is something between 0 and 1, and in your paper you also have sigmoid function in the supplementary material table 2. However I cannot find any sigmoid in the code; if the attention is without sigmoid in my opinion it is strange. Does it also work without sigmoid? Or did I miss anything?

您好，非常感谢您做出的卓越贡献。我在阅读您论文后(ECCV2020)，试着通过代码加深理解，然后发现在几个问题上有点疑惑，希望您可以帮忙解答。

1.您论文中提出了Pixel-wise以及Voxel-wise的视图选择模块，但是代码中只给出了后者，请问什么时候能将前者加入代码呢？

2.对于给出的Voxel-wise模块，对于第一个3D卷积，论文的Fig. 4中给的是3D Resblock，Table.1给出的是Conv3DGR，而在代码中，使用的是Conv3D。请问应该以哪个为准呢？

3.同样是Voxel-wise模块，论文里 Table.1给出的Conv3DGR中，使用的kernel size=3，而代码中都是使用kernel size=1，请问两者有什么区别或讲究嘛？这一块在论文中好像没有体现。或者哪个效果更好？

conv3d(in_channels, 1, kernel_size=1, stride=1, bias=bias),
conv3d(1, 1, kernel_size=1, stride=1)

4.想知道Issue3中提到的CTF是什么呢，论文里貌似没提到？

不好意思问题稍微有点多，希望您能抽出时间帮助我一下。谢谢您！

Hi, thank you very much for your great work and code.But when I use eval_pyramid.py on Tanks and Temples datasets,I meet this error:

inverse depth
inverse depth
process ['Family/{}/00000000{}']
torch.Size([1, 5, 3, 1064, 1920])
inverse depth
inverse depth
inverse depth
inverse depth
torch.Size([1, 128, 67, 120])
Traceback (most recent call last):
File "eval.py", line 366, in
save_depth()
File "eval.py", line 162, in save_depth
outputs = model(sample_cuda["imgs"], sample_cuda["proj_matrices"], sample_cuda["depth_values"])
File "/home/u/anaconda3/envs/neucon/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/home/u/anaconda3/envs/neucon/lib/python3.7/site-packages/torch/nn/parallel/data_parallel.py", line 153, in forward
return self.module(*inputs[0], **kwargs[0])
File "/home/u/anaconda3/envs/neucon/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/mnt/b7074bde-e07b-4bc7-8b24-a7b759401d85/PVAMVSNet-master/models/vamvsnet.py", line 567, in forward
features = [self.feature(img) for img in imgs]
File "/mnt/b7074bde-e07b-4bc7-8b24-a7b759401d85/PVAMVSNet-master/models/vamvsnet.py", line 567, in
features = [self.feature(img) for img in imgs]
File "/home/u/anaconda3/envs/neucon/lib/python3.7/site-packages/torch/nn/modules/module.py", line 722, in _call_impl
result = self.forward(*input, **kwargs)
File "/mnt/b7074bde-e07b-4bc7-8b24-a7b759401d85/PVAMVSNet-master/models/vamvsnet.py", line 172, in forward
x = torch.cat((conv5_0, conv3_2), dim=1)
RuntimeError: Sizes of tensors must match except in dimension 2. Got 133 and 134 (The offending index is 0)

How can I solve it?

Hi,Thanks a lot for the excellent code，But when I run eval_pyramid.sh I set fusion=True and I meet this error.How can I solve this?

eval.py:111: DeprecationWarning: np.bool is a deprecated alias for the builtin bool. To silence this warning, use bool by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use np.bool_ here.
Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
assert mask.dtype == np.bool
processing ./dtu/scan1, ref-view00, photo/geo/final-mask:0.5489484797297297/0.0005194256756756757/0.00035472972972972974
valid_points 0.00035472972972972974
Traceback (most recent call last):
File "eval.py", line 379, in
filter_depth(scan_folder, out_folder, os.path.join(save_dir, 'mvsnet{:0>3}_l3.ply'.format(scan_id)))
File "eval.py", line 330, in filter_depth
color = ref_img[1:-16:4, 1::4, :][valid_points] # hardcoded for DTU dataset
IndexError: boolean index did not match indexed array along dimension 0; dimension is 296 but corresponding boolean dimension is 592

您好，请问您如何可视化最终的点云数据，我用meshlab 打开后无法显示？

你好，非常感谢您的优秀工作和优质代码。
当我尝试利用tf_model中3DCNN中训练好的模型进行测试时出现 ./eval_pyramid.sh: line 22: ckpt: command not found，以及
Traceback (most recent call last):
File "eval.py", line 76, in
model_name = str.split(args.loadckpt, '/')[-2] + '_' + str.split(args.loadckpt, '/')[-1]
IndexError: list index out of range
这个问题，我是在readme的testing部分进行改动的，按照上面要求，我eval_pyramid.sh中将ckpt和checkpoint_list改成
checkpoint_list=(11)
ckpt = /media/smartcity/E6AA1145AA1113A1/LiSenTao/PVAMVSNet-master/MODEL_FOLDER/tf_model/GRU
我自己怀疑是这里的路径出了问题，但我自己感觉不知道改什么。最主要的是，我不是很明白checkpoint_list和ckpt后面到底需要加什么。