first commit

LICENSE.md

@@ -0,0 +1,9 @@
+The MIT License (MIT)
+
+Copyright (c) 2020
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

readme.md

@@ -0,0 +1,192 @@
+# NLOS-OT
+
+Official implementation of NLOS-OT: Passive Non-Line-of-Sight Imaging Using Optimal Transport (IEEE TIP, accepted)
+
+> Ruixu Geng, Yang Hu, Zhi Lu, Cong Yu, Houqiang Li, Hengyu Zhang and Yan Chen, “Passive Non-Line-of-Sight Imaging Using Optimal Transport,” IEEE Transactions on Image Processing, 2021
+
+```
+@article{nlosot,
+	title = {Passive Non-Line-of-Sight Imaging Using Optimal Transport},
+	journal = {IEEE Transactions on Image Processing},
+	author = {Geng, Ruixu and Hu, Yang and Lu, Zhi and Yu, Cong and Li, Houqiang and Zhang, Hengyu and Chen, Yan},
+	year = {2021}
+}
+```
+
+## Description
+
+In this repository, we release the ``NLOS-OT`` code in Pytorch as well as the passive NLOS imaging dataset ``NLOS-Passive``.
+
+- Problem statement: Passive NLOS imaging
+<p align="center"><img width="90%" src="pic/problem_statement.jpg" /></p>
+
+- NLOS-OT architecture
+<p align="center"><img width="90%" src="pic/NLOS-OT.jpg" /></p>
+
+- The reconstruction results of NLOS-OT trained by specific dataset without partial occluder
+<p align="center"><img width="90%" src="pic/results_specific.jpg" /></p>
+
+- The generalization results of NLOS-OT trained by dataset only from [STL-10](https://cs.stanford.edu/~acoates/stl10/) with unknown partial occluder 
+<p align="center"><img width="90%" src="pic/generalization.jpg" /></p>
+<p align="center"><img width="90%" src="pic/generalization_2.jpg" /></p>
+
+## Installation
+
+1. install required packages
+
+2. clone the repo
+
+
+## Prepare Data
+
+1. Download dataset
+
+You can download each group in NLOS-Passive through the link below. Please note that a compressed package (.zip or .z01+.zip) represents a group of measured data.
+
+link：https://pan.baidu.com/s/19Q48BWm1aJQhIt6BF9z-uQ 
+
+code：j3p2
+
+If the link fails, please feel free to contact me.
+
+> In the following content, we will take the partially occluded dataset on STL-10 (i.e., NLOS_Passive/STL-10/stl10_dark_1_d100_occluder.zip) as an example to illustrate the training and testing process. The dataset also contains test data from other data (MNIST, supermodel faces and real scenes), which can be used to evaluate the generalization ability of NLOS-OT.
+
+>> Similarly, you can also use other datasets to complete training and testing according to the content below.
+
+2. Organize the files structure of the dataset
+
+	+ Unzip the dataset
+
+	The size of a group projection dataset on STL-10 is about 30GB, which exceeds the maximum limit of Baiduyun. Therefore, we divided each group of data into two compressed packages (.z01 and .zip). You can decompress with
+
+	```bash
+	cd /pathtodataset/
+	zip -s 0 stl10_dark_1_d100_occluder.zip --out stl10_dark_1_d100_occluder_single.zip
+	unzip stl10_dark_1_d100_occluder_single.zip
+	# delete temp zip
+	rm -rf ./stl10_dark_1_d100_occluder_single.zip
+	# rename the folder
+	mkdir ./C_dark_1_d100_occluder/
+	mv ./pro ./C_dark_1_d100_occluder/train
+	```
+
+	The GT folder can be obtained by 
+	```bash
+	# Download gt zip. e.g., GT_stl10_allimages.zip
+	# unzip and rename
+	unzip GT_stl10_allimages.zip
+	```
+
+	+ Organize the dataset (rename and move)
+
+		We recommend organizing the data files according to the following structure:
+
+		```
+		.
+		├── B
+		│   └── train
+		│   	└── 1.png
+		│   	...
+		├── C_dark_1_d100_wall
+		│   └── train
+		│   	└── 1.png
+		│   	...
+		├── C_dark_2_d100_wall
+		│   └── train
+		│   	└── 1.png
+		│   	...
+		└── test 
+			├── C_dark_1_d100_wall_test
+			│   	└── 2.png
+			│   	...
+			├── C_dark_1_d100_wall_val
+			│   	└── 3.png
+			│   	...
+			├── C_dark_2_d100_wall_test
+			│   	└── 2.png
+			│   	...
+			├── C_dark_2_d100_wall_val
+			│   	└── 3.png
+			│   	...
+			├── gt_test
+			│   	└── 2.png
+			│   	...
+			└── gt_val
+			    	└── 3.png
+			    	...
+		```
+
+		In our paper, we organize the data files using the following commands:
+
+		+ For STL-10 (including real-world images)
+
+		+ For supermodel faces 
+
+		```bash
+		# gt
+
+		# projection images
+		mkdir ./test/C_dark_1_d100_wall_test
+		mkdir ./test/C_dark_1_d100_wall_val
+
+		mv ./C_dark_1_d100_wall/train/5* ./test/C_dark_1_d100_wall_val
+
+		mv ./test/C_dark_1_d100_wall_val/51* ./test/C_dark_1_d100_wall_test
+		mv ./test/C_dark_1_d100_wall_val/55* ./test/C_dark_1_d100_wall_test
+		mv ./test/C_dark_1_d100_wall_val/59* ./test/C_dark_1_d100_wall_test
+		```
+
+		The above structure can be applied to the training command we provided. You can also customize your own file structure and modify the corresponding parameters (--datarootTarget, --datarootData, --datarootValTarget, --datarootValData) in the command.
+
+## Demo / Evaluate
+
+Before that, you should have installed the required packages and organized the data set according to the appropriate file structure.
+
+1. Download pretrained pth
+
+2. run the test.py
+
+
+## Train
+
+Before that, you should have installed the required packages and organized the data set according to the appropriate file structure.
+
+
+## Citation
+- Feel free to use the dataset / code, but please cite:
+
+	- Ruixu Geng, Yang Hu, Zhi Lu, Cong Yu, Houqiang Li, Hengyu Zhang and Yan Chen, “Passive Non-Line-of-Sight Imaging Using Optimal Transport,” IEEE Transactions on Image Processing, 2021
+
+	```
+	@article{nlosot,
+		title = {Passive Non-Line-of-Sight Imaging Using Optimal Transport},
+		journal = {IEEE Transactions on Image Processing},
+		author = {Geng, Ruixu and Hu, Yang and Lu, Zhi and Yu, Cong and Li, Houqiang and Zhang, Hengyu and Chen, Yan},
+		year = {2021}
+	}
+	```
+
+- You may also be interested in our review article: 
+
+	- Ruixu Geng, Yang Hu, and Yan Chen, “Recent Advances on Non-Line-of-Sight Imaging: Conventional Physical Models, Deep Learning, and New Scenes,” APSIPA Transactions on Signal and Information Processing, 2021
+
+
+	If you think it is helpful, please consider citing
+
+	```
+	@article{RecentAdvancesNLOS,
+		author    = {Geng, Ruixu and Hu, Yang and Chen, Yan},
+		title     = {Recent Advances on Non-Line-of-Sight Imaging: Conventional Physical Models, Deep Learning, and New Scenes},
+		journal   = {APSIPA Transactions on Signal and Information Processing},
+		year      = {2021}
+	}
+	```
+
+
+- We thank the following great works:
+
+	- [DeblurGAN](https://github.com/KupynOrest/DeblurGAN), [pix2pix](https://github.com/phillipi/pix2pix): Our code is based on the framework provided by the two repos.
+
+	- [IntroVAE](https://proceedings.neurips.cc/paper/2018/hash/093f65e080a295f8076b1c5722a46aa2-Abstract.html): The encoder and decoder in NLOS-OT are based on IntroVAE.
+
+	- [AE-OT](https://openreview.net/forum?id=HkldyTNYwH), [AEOT-GAN](https://arxiv.org/abs/2001.03698): The idea of using OT to complete passive NLOS imaging tasks in NLOS-OT comes from the two works.

step1/INstructor.txt

@@ -0,0 +1 @@
+conda activate pytorch17; CUDA_VISIBLE_DEVICES=7 pytnon train.py --datarootTarget '/share2/data/AONLOS/STL10/B/' --datarootData '/share2/data/AONLOS/STL10/B/' --datarootValTarget '/share2/data/AONLOS/STL10/test/gt_val/' --datarootValData '/share2/data/AONLOS/STL10/test/gt_val/'  --learn_residual --gpu_ids='0' --batchSize=16 --name='NEWRight_transformer_stl10_batch_Tanh' --niter=100 --niter_decay=300 --which_model_netG transformer --display_port=9099 --norm 'batch'

step1/LICENSE

@@ -0,0 +1,58 @@
+Copyright (c) 2017, Jun-Yan Zhu and Taesung Park
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
+
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+
+--------------------------- LICENSE FOR pix2pix --------------------------------
+BSD License
+
+For pix2pix software
+Copyright (c) 2016, Phillip Isola and Jun-Yan Zhu
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
+
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+
+----------------------------- LICENSE FOR DCGAN --------------------------------
+BSD License
+
+For dcgan.torch software
+
+Copyright (c) 2015, Facebook, Inc. All rights reserved.
+
+Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
+
+Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
+
+Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.
+
+Neither the name Facebook nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

step1/data/base_data_loader.py

@@ -0,0 +1,14 @@
+
+class BaseDataLoader():
+    def __init__(self):
+        pass
+
+    def initialize(self, opt):
+        self.opt = opt
+        pass
+
+    def load_data():
+        return None
+
+
+

step1/data/base_dataset.py

@@ -0,0 +1,15 @@
+import torch.utils.data as data
+from PIL import Image
+import torchvision.transforms as transforms
+import pdb
+
+class BaseDataset(data.Dataset):
+    def __init__(self):
+        super(BaseDataset, self).__init__()
+
+    def name(self):
+        return 'BaseDataset'
+
+    def initialize(self, opt):
+        pass
+

step1/data/custom_dataset_data_loader.py

@@ -0,0 +1,37 @@
+import torch.utils.data
+from data.base_data_loader import BaseDataLoader
+import pdb
+
+def CreateDataset(opt):
+    dataset = None
+    if opt.dataset_mode == 'single':  # yes
+        from data.single_dataset import SingleDataset   
+        dataset = SingleDataset()
+    elif (opt.dataset_mode == 'aligned') or (opt.dataset_mode == 'aligned'):
+        raise ValueError('In NLOS-OT, we only support dataset in single mode.')
+    else:
+        raise ValueError("Dataset [%s] not recognized." % opt.dataset_mode)
+
+    print("dataset [%s] was created" % (dataset.name()))    
+    dataset.initialize(opt)     
+    return dataset 
+
+
+class CustomDatasetDataLoader(BaseDataLoader):
+    def name(self):
+        return 'CustomDatasetDataLoader'
+
+    def initialize(self, opt):
+        BaseDataLoader.initialize(self, opt) 
+        self.dataset = CreateDataset(opt)
+        self.dataloader = torch.utils.data.DataLoader(
+            self.dataset,
+            batch_size=opt.batchSize,
+            shuffle=not opt.serial_batches,
+            num_workers=int(opt.nThreads))
+
+    def load_data(self):
+        return self.dataloader
+
+    def __len__(self):
+        return min(len(self.dataset), self.opt.max_dataset_size)

step1/data/data_loader.py

@@ -0,0 +1,7 @@
+
+def CreateDataLoader(opt):
+    from data.custom_dataset_data_loader import CustomDatasetDataLoader
+    data_loader = CustomDatasetDataLoader()
+    print(data_loader.name())
+    data_loader.initialize(opt)
+    return data_loader