geozcx / a-deeply-supervised-image-fusion-network-for-change-detection-in-remote-sensing-images Goto Github PK

模型权重的链接挂了，可以再上一下吗，感谢

I superimposed the two images into an array and imported the model. Always report errors as follows:

Input 0 of layer block1_pool is incompatible with the layer: expected ndim=4, found ndim=5. Full shape received: [4, 2, 512, 512, 64]

This is my code: model.fit(train_imag,train_label,batch_size = 4,epochs=100)

`def traindataloard(path1, path2):
img1path = os.listdir(path1)
img2path = os.listdir(path2)
img1group = [input_t1 + name1 for name1 in img1path]
img2group = [input_t2 + name2 for name2 in img2path]
train_img2 = []
for img1n, img2n in zip(img1group, img2group):
train_img1 = []
img1 = cv2.imread(img1n)
img2 = cv2.imread(img2n)

    img1 = np.array(img1)  
    img2 = np.array(img2)

    train_img = [img1, img2]    
    train_img2.append(train_img)
return train_img2`

train_img = traindataloard(input_t1, input_t2)

In the pytorch version code:

1. 'Sigmoid' is used when you build DSIFN model, but it's used again in your loss function.

2. Then, in loss.py, why use 'Sigmoid' function for 'target' ?

I want to do some research on change detection in big map. But there are few public change detection dataset of big map. I see the raw big maps are shown in readme. Does it mean that the raw image can be obtained by stitching, or the raw image will be made public in the future?

Hi @GeoZcx, I find a phenomenon that when I initialization of network parameters, the network can't detect any change in Pytorch Version. But when I don't initialize of network parameters the network works well. The follow is the code of initialization

`
def _initialize_weights(self):

    for m in self.modules():
        if isinstance(m, nn.Conv2d) or isinstance(m, nn.ConvTranspose2d) or isinstance(m, nn.Linear):
            nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            if m.bias is not None:
                m.bias.data.zero_()
        elif isinstance(m, nn.BatchNorm2d):
            m.weight.data.fill_(1)
            m.bias.data.zero_()

`

For citing and fair comparisons, would you mind providing the pre-trained weights of the DSIFN and the predicted change maps used in the paper?

I am a beginner. I have programed the train code and the dataloader function about the dataset, and I have set the batchsize of 32, but it still out of memories, could u please tell the way to solve the memory problem?

The 512x512 resultion masks in the train and val set that you released are all black. Did you train your model by upsampling the masks that resultion is 256x256, or some other methods ?

In Pytorch Version, only have one Spatial_Attention，but I think there be 5 spatial_attentions is beeter. Because there are 5 branch, and each branch should have it's own spatial_attention after channel attention. And I found there are 5 spatial_attentions in Keras Version.

Thanks your new idea. There is a discrepancy with the explanation of the paper in Pytorch Version.

Hello, I am very interested in your paper, can re-upload the weight files? it looks like already invalidated

I found that the loss funcion cannot handle the tuple returned by your model. The code of the PyTorch version may wrong.

The loss may like the following formats

for itr in itr_loss: loss += F.binary_cross_entropy(input, target)

Hi, I am confused about your training process. Based on my understanding，you load 'imagenet' pretrained weights on Vgg16, then you freeze Vgg16 parameters and only train DDN part. Is my understanding right？According to the paper, I noticed that the deep feature extraction network (DFEN) and the difference discrimination network (DDN) in the proposed method are independently trained during the training process, but if you just load 'imagenet' pretrained weights on Vgg16, you don't train Vgg16. In addition to this, if you just use pretrained weights, I question the effect of pretrained weights.

作者您好，请问数据集mask文件全黑怎么解决呢

Sorry, my English is not very good. About your codes in pytorch version, I am comfused about a problem need your help. Why is the convolution output of the last layer of the network single channel? In remote sensing change detection. Generally, there are two types of change and non change. If output only single channel, how do we get the results of change?

Thank you for your generous sharing！

I found that the label of the CDD dataset was not binary map. How did you deal with this situation ?

thanks

During training, VGG backbone is frozen with pretrained weights by self.features = nn.ModuleList(features).eval() so that only the weight parameters of DDN will be updated.

Environment:

Ubuntu 16.04
Python==3.6.8
PyTorch==1.4
a RTX2080Ti GPU

Hyperparameters setting:

batch_size = 8
initial learning rate=0.0001
adam optimizer

Would you mind releasing the complete training code? Thanks.

Can you share the trianing.py and predict.py of troch version. I will appreciate you very much.

Can you share its Demo Or some Model life like I dont know how to use these code i have tried but failed and end up to make a Undesired output plzzzz help

https://github.com/likyoo/open-cd/tree/main/projects/open-cd_technical_report

Excuse me, i have a problem about the loss. the problem is that the output pixel of the branch is smaller than the label (such is the branch output is 2828,but the label is 256256),so how to deal with the label to fit the branch output pixel? i want to use downsampling...
looking foward to your reply!

作者你好，在公开代码中的DSFIN脚本中，不太懂model_A,model_B,是用来做什么的，请问有什么方式可以获得完整的代码吗？万分感谢

是否能够分享您的训练，验证，测试的源码，万分感谢

I want to know what batchsize is set to

When defining the model，generally this is the form："class Unet(nn.Module): def init(self, input_nbr, label_nbr):". Why is there no choice of input and output channels in your Pytorch codes.
In your codes, “class DSIFN(nn.Module): def init(self, model_A, model_B):”, model_A and model_B represent what? vgg16_base? ?
Looking forward to your answer, it will be very helpful to me. Thank you!

Hello, I recently came across your paper. I am unable to find the training code for your model. Can you please share your training code for keras? Can you provide it here or in the google drive link for the dataset?
Thank you!

请问预训练模型跑出的predicted image为什么会和论文里给出的效果差很多

Hi @GeoZcx
First of all thanks for your great repo.

Unfortunately, Beidu drive isn't available in my country.
Could you please provide a Google drive or even Onedrive version of it?

During training, VGG backbone is frozen with pretrained weights by self.features = nn.ModuleList(features).eval() so that only the weight parameters of DDN will be updated.

Environment:

Ubuntu 16.04
Python==3.6.8
PyTorch==1.4
a RTX2080Ti GPU

Hyperparameters setting:

batch_size = 8
initial learning rate=0.0001
adam optimizer

Would you mind releasing the complete training code? Thanks.

Hello, I am very interested in your paper, but I have some problems in understanding your paper and code. Your paper said that the change detection task is divided into two stages. First, the bitemporal image is sent to the pre-trained DFEN (two VGG networks) to generate deep features, and then the extracted deep features are input into DDN to distinguish the changed areas. What do you mean by saying that these two parts should be trained independently? Does it mean to use the feature extraction method in transfer learning to extract features and then directly put them into DDN for training? You did not use the fine-tuning method, right? Hope to hear from you , thank you very much!

Hello, what is the role of model_A and model_B or where is its definition, and what should be input? What did they do to "t1_input" and "t2_input" ? Thanks.

Hello, I did not find code for training a model in this thesis code of yours

感谢您的工作
您在论文使用了CDD数据集，并在其测试集上取得了很好的结果。我想问一下您在计算评估指标的时候，比如用F1-Sore，那应该求出测试集中每张图片（256*256）计算出F1，再求平均；还是求出整个测试集的混淆矩阵，再算F1-Score呢？

Hi,
Thank you for sharing code.
Can you share the train code, I am a complete beginner, and what does the f1 mean for in the load_model fun?
thank you so much!