Super-Resolution Network

Our PyTorch implementation of [1], please visit their repository for more details.

Content

• Installation
• Data
• Super-Resolution
• Sources

Installation

conda create -n super-resolution-pytorch
conda activate super-resolution-pytorch
conda env update -n super-resolution-pytorch --file environment.yml

Data

The ShapeNet dataset can be found at [2] and must be placed in the directory data/ShapeNetCorev1.

Super-Resolution

Please be sure, that the data is inside the respective directory. For training the Super-Resolution-Network, first pre-process the data with:

python prepare_SR.py

A new folder with the different voxel grid resolutions and ODMs will be created in data/ShapeNetCoreSR. If you want, you can remove the folder data/ShapeNetCorev1 manually.

Then train the depth map model with:

python train_SR.py --model_type "depth"

Followed with the occupancy map model:

python train_SR.py --model_type "occupancy"

If you use our PyTorch implementation, please refer to our repository https://github.com/mhohmann1/super-resolution-pytorch and of course to [1].

Sources

[1] https://github.com/EdwardSmith1884/Multi-View-Silhouette-and-Depth-Decomposition-for-High-Resolution-3D-Object-Representation

[2] ShapeNetCorev1 - https://shapenet.org/