[CVPR 2020] 3D Photography using Context-aware Layered Depth Inpainting

This is a modified version for Windows use. All credits still belong to the original researchers.

[Paper] [Project Website] [Google Colab]

We propose a method for converting a single RGB-D input image into a 3D photo, i.e., a multi-layer representation for novel view synthesis that contains hallucinated color and depth structures in regions occluded in the original view. We use a Layered Depth Image with explicit pixel connectivity as underlying representation, and present a learning-based inpainting model that iteratively synthesizes new local color-and-depth content into the occluded region in a spatial context-aware manner. The resulting 3D photos can be efficiently rendered with motion parallax using standard graphics engines. We validate the effectiveness of our method on a wide range of challenging everyday scenes and show fewer artifacts when compared with the state-of-the-arts.

3D Photography using Context-aware Layered Depth Inpainting
Meng-Li Shih, Shih-Yang Su, Johannes Kopf, and Jia-Bin Huang
In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020.

Prerequisites

• Anaconda3

You can follow a video tutorial here if it's more convenient for you.

• To get started, please run the following steps:

• Clone this page

• Clone this page

  Unzip, drag and drop it here, follow the below folder structure(change the file names too):

  📂3d-photo-inpainting/
  ├── 📂BoostingMonocularDepth/ <--
  │...
  

Change your Anaconda directory to that folder with cd YOUR_FILE_DIRECTORY/3d-photo-inpainting/

• Next, follow these installation steps under Anaconda Prompt:

  conda create -n 3DP python=3.7
  conda activate 3DP

  For any other GPU that's < RTX 30 series:

  conda install pytorch==1.5.0 torchvision==0.6.0 cudatoolkit=10.2 -c pytorch

  For any RTX 30 series:

  conda install pytorch torchvision torchaudio cudatoolkit=11.1 -c pytorch -c conda-forge

  Continue:

  pip install decorator
  pip install -r requirements.txt
  pip install pyyaml
  pip install cython
  pip install pyqt5
  pip install matplotlib

  If you cannot install Cython, cynetworkx successfully, please install Git for Windows link

  And then run the following commands:

  pip install Cython==3.0a5

  pip install git+https://github.com/pattern-inc/cynetworkx.git

• Next, please download the following models & weights and put them in the right file directory (from my backup IceDrive):

  • color-model.pth
  • depth-model.pth
  • edge-model.pth
  • model.pt
  • model-f46da743.pt
  • latest_net_G.pth

  📂3d-photo-inpainting/
  ├── 📂checkpoints/
  │   ├── 📜color-model.pth
  │   ├── 📜depth-model.pth
  │   └── 📜edge-model.pth
  ├── 📂MiDaS/
  │   └── 📜model.pt
  ├── 📂BoostingMonocularDepth/
  │   ├── 📂midas/
  │   │   └── 📜model-f46da743.pt (rename to model.pt after dragged in)
  │   └── 📂pix2pix/
  │       └── 📂checkpoints/
  │           └── 📂mergemodel/
  │               └── 📜latest_net_G.pth
  │...
  

Quick start

Please follow the instructions in this section. This should allow to execute our results. For more detailed instructions, please refer to DOCUMENTATION.md.

Execute

1. Put .jpg files (e.g., test.jpg) into the image folder. Don't use any space in the image name (eg. "image 1.jpg" would not work)
   • E.g., image/moon.jpg
2. Run the following command

   python main.py --config argument.yml

   • Note: The 3D photo generation process usually takes about 2-3 minutes depending on the available computing resources.
3. The results are stored in the following directories:
   • Corresponding depth map estimated by MiDaS
     • E.g. depth/moon.npy, depth/moon.png
     • User could edit depth/moon.png manually.
       • Remember to set the following two flags as listed below if user wants to use manually edited depth/moon.png as input for 3D Photo.
         • depth_format: '.png'
         • require_midas: False
   • Inpainted 3D mesh (Optional: User need to switch on the flag save_ply)
     • E.g. mesh/moon.ply
   • Rendered videos with zoom-in motion
     • E.g. video/moon_zoom-in.mp4
   • Rendered videos with swing motion
     • E.g. video/moon_swing.mp4
   • Rendered videos with circle motion
     • E.g. video/moon_circle.mp4
   • Rendered videos with dolly zoom-in effect
     • E.g. video/moon_dolly-zoom-in.mp4
     • Note: We assume that the object of focus is located at the center of the image.
4. (Optional) If you want to change the default configuration. Please read DOCUMENTATION.md and modified argument.yml.

License

This work is licensed under MIT License. See LICENSE for details.

If you find our code/models useful, please consider citing our paper:

@inproceedings{Shih3DP20,
  author = {Shih, Meng-Li and Su, Shih-Yang and Kopf, Johannes and Huang, Jia-Bin},
  title = {3D Photography using Context-aware Layered Depth Inpainting},
  booktitle = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year = {2020}
}

Acknowledgments

• We thank Pratul Srinivasan for providing clarification of the method Srinivasan et al. CVPR 2019.
• We thank the author of Zhou et al. 2018, Choi et al. 2019, Mildenhall et al. 2019, Srinivasan et al. 2019, Wiles et al. 2020, Niklaus et al. 2019 for providing their implementations online.
• Our code builds upon EdgeConnect, MiDaS and pytorch-inpainting-with-partial-conv