can you share the dataset for change detection

[ARnnn]

can you share the dataset for change detection

[ARnnn]

1:When I read the metrics.py ,I get two mean_iou(),which is right, please?
2:I found this data set, but it's a four-channel one. Did you take three of them to do it?

[ARnnn]

How many pictures did you use in training?

[Hhhhhhao]

@ARnnn Hi, sorry for the delayed response.
The dataset I used is a change detection dataset from here: http://rscup.bjxintong.com.cn/.
The second mean_iou function is better.
For the dataset I used 4 channels directly as input.

[ARnnn]

thanks for you reply.ang in addition ,in your test.py file, where are val_img_paths, val_mask_paths defined, and where are the test data to be passed in?the most important is that how can i run the test.py

[ARnnn]

thanks for you reply.ang in addition ,in your test.py file, where are val_img_paths, val_mask_paths defined, and where are the test data to be passed in?the most important is that how can i run the test.py

[ARnnn]

thanks for you reply.ang in addition ,in your test.py file, where are val_img_paths, val_mask_paths defined, and where are the test data to be passed in?the most important is that how can i run the test.py

[Hhhhhhao]

@ARnnn The dataset.py file directly read the data, so long as you are putting the data under the data folder.

[ARnnn]

but when i run the test.py ,i got the error:
=> creating model NestedUNet
Traceback (most recent call last):
File "test.py", line 104, in
main()
File "test.py", line 48, in main
model = models.dictargs.arch
File "I:\change_detection-master\change_detection-master\models.py", line 71, in init
self.conv0_0 = ResBlock(in_ch, filters[0])
File "I:\change_detection-master\change_detection-master\models.py", line 43, in init
self.conv1 = nn.Conv2d(in_channels=in_ch, out_channels=out_ch, kernel_size=3, stride=stride, padding=1, dilation=1, bias=True)
File "D:\Anaconda3\lib\site-packages\torch\nn\modules\conv.py", line 327, in init
False, _pair(0), groups, bias, padding_mode)
File "D:\Anaconda3\lib\site-packages\torch\nn\modules\conv.py", line 21, in init
if in_channels % groups != 0:
TypeError: unsupported operand type(s) for %: 'Namespace' and 'int'

how can i solve it

[Hhhhhhao]

@ARnnn I think there is a bug

line 48 should read arg.channels into model rather than arg.
I fixed it.

[GH-W5]

1：当我读取metrics.py时，我得到两个mean_iou（），对吗？
2：我找到了这个数据集，但这是一个四通道数据集。您是否接受了其中的三个？

Can you share the dataset?

[ARnnn]

in test.py,what do val_img_paths and val_mask_paths represent and where are they defined?

[Hhhhhhao]

the dataset can be download from here: http://rscup.bjxintong.com.cn/#/theme/4

the val_ima_paths and mask_paths are defined in dataset.py

[ARnnn]

I'm talking about val_img_paths on line 71 and val_mask_paths on line 80 in test.py. As you know, I didn't find them in dataset.py. Can you tell me the number of rows in dataset.py?

[ARnnn]

My problem now is that when I run test.py, the following errors occur:

Traceback (most recent call last):
File "test.py", line 107, in
main()
File "test.py", line 80, in main
for i in tqdm(range(len(val_mask_paths))):
NameError: name 'val_mask_paths' is not defined

how can i solve it?

[ARnnn]

I think that 79 to 99 lines of code in test. py are an evaluation criterion, so comment it out. ThenI run test.py, the following errors occur:

Traceback (most recent call last):
File "test.py", line 116, in
main()
File "test.py", line 73, in main
for i, (input, target) in tqdm(enumerate(val_loader), total=len(val_loader))
:
File "D:\Anaconda3\lib\site-packages\tqdm_tqdm.py", line 1005, in iter
for obj in iterable:
File "D:\Anaconda3\lib\site-packages\torch\utils\data\dataloader.py", line 346
, in next
data = self.dataset_fetcher.fetch(index) # may raise StopIteration
File "D:\Anaconda3\lib\site-packages\torch\utils\data_utils\fetch.py", line 4
7, in fetch
return self.collate_fn(data)
File "E:\z\change_detection-master\change_detection-master\data_loaders.py", l
ine 11, in collate_fn
img, mask = data[0]
ValueError: too many values to unpack (expected 2)

how can i solve it?

[ARnnn]

Is the data you used to test a test set of 83 pictures provided by the competitors?

[ARnnn]

How do you make 256 * 256 pictures for training? How many pictures are used for training

[ARnnn]

Why is the test result all black after I run with your code?

[Hhhhhhao]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

[ARnnn]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

[Hhhhhhao]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

[ARnnn]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

I can give you the original training set. Can you train again? esteem it a favor!

[ARnnn]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

I can give you the original training set. Can you train again? esteem it a favor!

I really want to reproduce this code.

[Hhhhhhao]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

I can give you the original training set. Can you train again? esteem it a favor!

I really want to reproduce this code.

really want to help but don't have GPU currently

[ARnnn]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

I can give you the original training set. Can you train again? esteem it a favor!

I really want to reproduce this code.

really want to help but don't have GPU currently

Can you tell me how you organized the original training data into 256 sizes in the early stage? What are the details? In addition, what was your operating environment at that time?

[deimsdeutsch]

Unable to download dataset ... can you share baidu or google drive links ?

[Rikaaaaaa95]

@ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training?

I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model?

sorry I didn;t have the data and model any more

I can give you the original training set. Can you train again? esteem it a favor!

I really want to reproduce this code.

really want to help but don't have GPU currently

Can you tell me how you organized the original training data into 256 sizes in the early stage? What are the details? In addition, what was your operating environment at that time?

可以留一下qq么，在很多change detection下看到你

[Hhhhhhao]

留下邮箱吧 把QQ发你

…

On Jan 23, 2020, at 09:02, Rikaaaaaa95 ***@***.***> wrote:  @ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training? I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model? sorry I didn;t have the data and model any more I can give you the original training set. Can you train again? esteem it a favor! I really want to reproduce this code. really want to help but don't have GPU currently Can you tell me how you organized the original training data into 256 sizes in the early stage? What are the details? In addition, what was your operating environment at that time? 可以留一下qq么，在很多change detection下看到你 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

[Rikaaaaaa95]

留下邮箱吧 把QQ发你
…
On Jan 23, 2020, at 09:02, Rikaaaaaa95 @.***> wrote:  @ARnnn maybe you have too much black pixels in your ground truth than the white pixels. Have you checked the detection masks generated during training? I cut the original image into 256 small pieces to train, and ensure that every training set has white pixels, but after training, the test model is all black. Can you upload your training set and model? sorry I didn;t have the data and model any more I can give you the original training set. Can you train again? esteem it a favor! I really want to reproduce this code. really want to help but don't have GPU currently Can you tell me how you organized the original training data into 256 sizes in the early stage? What are the details? In addition, what was your operating environment at that time? 可以留一下qq么，在很多change detection下看到你 — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.

690954421@qq.com

[Rikaaaaaa95]

Unable to download dataset ... can you share baidu or google drive links ?

你好，能留下联系方式么，我也没下载好数据集

[Rikaaaaaa95]

can you share the dataset for change detection

请问你的qq是多少？