hazelu1/mogan

MOGAN

Official pytorch implementation of the paper: "MOGAN: Morphologic-structure-aware Generative Learning from a Single Image"

Random image generation for ROI-based tasks

Given ROIs, MOGAN is able to generate samples with diverse appereances and reasonable structures.

Link

Arxiv

Usage

Main Dependencies

matplotlib
scikit-image
scikit-learn
scipy
numpy
torch
torchvision

Training

python run.py --mode f --config_file test.config

TODO list:

• Clean useless comments.
• PEP8.
• Add full applications.
• Add the code for image fusing.
• Add more detailed instructions.

Note that

The code is incomplete for now. But the code of model's structure and training process is avaliable. Experiments haven't been stoped on a more reliable structure design so the code in the repo shouldn't be the final version.