train on ETH3D dataset

[wangtao-hku]

Hi,

I downloaded the ETH3D dataset from MVSNET repo and use 'convet_eth3d_dataset.py' to process it.

However, i encountered an error during the training process.

Traceback (most recent call last):
File "c:\Users\SmartLab\Desktop\dl_recon_semantics\codes\Recon\PatchmatchNet-main\train.py", line 313, in
train(input_args, pmnet_model, train_loader, test_loader, pmnet_optimizer, epoch_start)
File "c:\Users\SmartLab\Desktop\dl_recon_semantics\codes\Recon\PatchmatchNet-main\train.py", line 41, in train
process_samples(args, train_sample, "train", logger, model, train_image_loader, optimizer, epoch_idx)
File "c:\Users\SmartLab\Desktop\dl_recon_semantics\codes\Recon\PatchmatchNet-main\train.py", line 92, in process_samples
for batch_idx, sample in enumerate(image_loader):
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 652, in next
data = self._next_data()
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 1347, in _next_data
return self._process_data(data)
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 1373, in _process_data
data.reraise()
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch_utils.py", line 461, in reraise
raise exception
ValueError: Caught ValueError in DataLoader worker process 0.
Original Traceback (most recent call last):
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data_utils\worker.py", line 302, in _worker_loop
data = fetcher.fetch(index)
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "C:\Users\SmartLab\anaconda3\lib\site-packages\torch\utils\data_utils\fetch.py", line 49, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "c:\Users\SmartLab\Desktop\dl_recon_semantics\codes\Recon\PatchmatchNet-main\datasets\mvs.py", line 95, in getitem
depth_gt = read_map(depth_gt_filename, self.max_dim).transpose([2, 0, 1]).copy()
ValueError: axes don't match array

any help is much appreciated.

[HiRenZeHui]

Hello, I also encountered this problem. Have you solved it

[FangjinhuaWang]

I did not try with training on eth3d. But seems like the problem is that the array only has two dimension, so it pops out error when doing 'transpose([2, 0, 1])'

[LIN-FZ]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

[LIN-FZ]

Hello, I also encountered this problem. Have you solved it

Hello, have you successfully used ETH3D training patchmatchnet?

[wangtao-hku]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Hello, I also encountered this problem. Have you solved it

I did not proceed with it. ETH3D is gray and the dimension is not divisible by 8 if I remember correctly while Patchmatchnet accepts RGBs with dimension divisible by 8

[wangtao-hku]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

[She-Zhao]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

[wangtao-hku]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

[She-Zhao]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

How do you train? Can we contact in another way (such as email or wechat)? Thank you very much！

[wangtao-hku]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

How do you train? Can we contact in another way (such as email or wechat)? Thank you very much！

you can send me your email or WeChat and I will contact you,

[ym19196018395]

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

How do you train? Can we contact in another way (such as email or wechat)? Thank you very much！

you can send me your email or WeChat and I will contact you,

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

In that case, can we only use the DTU dataset for training? Are there other existing data sets that can be used to train new networks.

Can try BlendedMVS

I am trying BlendedMVS，and I get the same problem as you.I checked the pfm format of the depth map and found that it was grayscale rather than rgb.

BlendedMVS is fine, I trained on BlendedMVS with no problem.

How do you train? Can we contact in another way (such as email or wechat)? Thank you very much！

you can send me your email or WeChat and I will contact you,

@TWang1017 Hello author, could I ask you how to train on BlendedMVS data? Thank you.