chn0318/recurrent-denoising-autoencoder

Recurrent-Denoising-Autoencoder

Link to original paper (SIGGRAPH '17) : https://research.nvidia.com/publication/interactive-reconstruction-monte-carlo-image-sequences-using-recurrent-denoising

note: This is an unofficial PyTorch implementation of the above paper.

I also refer to the following links：

AakashKT/pytorch-recurrent-ae-siggraph17: Pytorch implementation for 'Interactive Reconstruction of Monte Carlo Image Sequences using a Recurrent Denoising Autoencoder' ;

Requirements

• Python >= 3.5 (3.7 recommended)
• torchvision
• numpy
• tensorboard
• pandas
• pillow
• matplotlib
• openCV

To install other dependencies, you can use pip with :

pip install -r requirements.txt

Demo

The original video:

Video after noise reduction:

Folder Structure

The folder structure of the project is shown below:

├─Recurrent-Denoising-Autoencoder
  │  commend.txt                                        
  │  data_chn.py                                         
  │  example.png                                         
  │  grab_frame.py                                      
  │  losses_chn.py                                       
  │  measure.py                                          
  │  model_chn.py                                        
  │  test.py                                           
  │  train.py                                             
  │
  ├─logs_train                                            
  │      events.out.tfevents.1639969697.LAPTOP-HHJKMK56.22388.0
  │
  ├─output                                                
  │  └─seq_1
  │          1.png
  │          2.png
  │          3.png
  │          4.png
  │          5.png
  │          6.png
  │          7.png
  │
  ├─save                                                  
  │      chn_test_201.pt
  │
  ├─test_data                                  
  │  └─seq_1
  │          1.png
  │          2.png
  │          3.png
  │          4.png
  │          5.png
  │          6.png
  │          7.png
  │
  └─__pycache__
          data_chn.cpython-37.pyc
          losses_chn.cpython-37.pyc
          model_chn.cpython-37.pyc

Dataset

The input to this network is :

• noisy input（1spp RGB image）
• target image
• Albedo
• Depth
• Normal Map
• Roughness Map

Construct the input as one image, as follows :

The directory structure of train data:

• [DATA_DIRECTORY]/train]
  • [DATA_DIRECTORY]/train/seq_0/
  • [DATA_DIRECTORY]/train/seq_1/
  • ......

There are 7 continuous frames in each [DATA_DIRECTORY]/train/seq_i

Train

To train the network, run the following command :

python train.py --name [EXP_NAME] --data_dir  [PATH_TO_DATA_DIRECTORY]  --save_dir [PATH_TO_SAVE_CHECKPOINTS] --width [width of image] --height [height of image] --epochs 100

Test

To test the network, run the following command :

python test.py  --test_dir  [PATH_TO_DATA_DIRECTORY]   --output_dir [PATH_TO_SAVE_RESULTS]  --checkpoint [PATH_TO_CHECKPOINT].pt --width [width of image] --height [height of image]