search

ZHU-Zhiyu / PZRes-Net

Source Codes of the paper " Hyperspectral Image Super-resolution via Deep Progressive Zero-centric Residual Learning"
MIT License
52 stars 9 forks source link

The dataset preprocess method not publish #1

Open erqihuang opened 3 years ago

erqihuang commented 3 years ago

In the Hyper_dataset class has used many data file, which the author does not provide the process code or example one item. Please provide above it?

ZHU-Zhiyu commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

erqihuang commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

ZHU-Zhiyu commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

erqihuang commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

ZHU-Zhiyu commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

erqihuang commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me.

The PSNR during the training: image

The splitted of the training and testing set: image

ZHU-Zhiyu commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me. image

In our paper, we give the detailed training and testing samples in Figure.4. Plase train our network with illustration of Figure.4. Thanks.

erqihuang commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me. image

In our paper, we give the detailed training and testing samples in Figure.4. Plase train our network with illustration of Figure.4. Thanks.

After comparing the training samples, I strictly followed the paper guidance. Though the result has got the significant improvement, I still did not get the result as well as the paper? It's hopeful that I can get the more precisely method about the training network from you, please~

Test's PSNR during the training: image

Data division: image image

ZHU-Zhiyu commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me. image

In our paper, we give the detailed training and testing samples in Figure.4. Plase train our network with illustration of Figure.4. Thanks.

After comparing the training samples, I strictly followed the paper guidance. Though the result has got the significant improvement, I still did not get the result as well as the paper? It's hopeful that I can get the more precisely method about the training network from you, please~

Test's PSNR during the training: image

Data division: image image

Thanks very much for your interests on our project. I have corrected the code, where I have commented out the ensential component, zero-norm. You can try agin. If any questions, feel free to ask.

erqihuang commented 3 years ago

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me. image

In our paper, we give the detailed training and testing samples in Figure.4. Plase train our network with illustration of Figure.4. Thanks.

After comparing the training samples, I strictly followed the paper guidance. Though the result has got the significant improvement, I still did not get the result as well as the paper? It's hopeful that I can get the more precisely method about the training network from you, please~ Test's PSNR during the training: image Data division: image image

Thanks very much for your interests on our project. I have corrected the code, where I have commented out the ensential component, zero-norm. You can try agin. If any questions, feel free to ask.

Thanks for the question :), the source data of CAVE dataset is aviliable in https://drive.google.com/drive/folders/1rjvP9d0tmM_zkbXWfIMK4fDP6HVSMVI9?usp=sharing . We will provide source code on the other datasets soonly, after finishing some projects.

Thank for your sharing! But it make me a little confused that the SOTA experiment result did not achieve in which the input was CAVE x8 downsample HSI and the used network was spenet. Could you give me guidance which tell me how to train correctly?

Did you correctly divide the traing/testing dataset according to our paper? Different data division method may affect the performance.

Thank for your replying~ I chose the first twelve image as train set, and the other as test set. In the paper, I noticed that the data is randomly divided into train and test set. Could you tell me how to select the train and test set then the result being SOTA?

Please refer to our paper for data division. We give the detialed images used for training and testing. Moreover, we will also provide correspoding code soonly.

I splitted the training and testing test carefully according to reference in the paper(MHF-Net). After 1000 epoch training, I didn't get the SOTA result on testing set. I wished I got the SOTA result if the authors told the correct way to me. image

In our paper, we give the detailed training and testing samples in Figure.4. Plase train our network with illustration of Figure.4. Thanks.

After comparing the training samples, I strictly followed the paper guidance. Though the result has got the significant improvement, I still did not get the result as well as the paper? It's hopeful that I can get the more precisely method about the training network from you, please~ Test's PSNR during the training: image Data division: image image

Thanks very much for your interests on our project. I have corrected the code, where I have commented out the ensential component, zero-norm. You can try agin. If any questions, feel free to ask.

Thank for your replying! There is possible that you provide me with the trained model so that I can compare with the local model parameters.

ZHU-Zhiyu commented 3 years ago

There is possible that you provide me with the trained model so that I can compare with the local model parameters.

OK, here are the pretrained model with code. it can achieve SOTA performance: https://drive.google.com/drive/folders/12EF862NiA7enYze7HYazSioeIUVl51__?usp=sharing image