Skip to content

The official pytorch implementation of "Increasing diversity of omni-directional images generated from single image using cGAN based on MLPMixer" in ACPR2023

Notifications You must be signed in to change notification settings

islab-sophia/odigen-mlpmixer

Repository files navigation

#odigen-mlpmixer

This repository contains the official pytorch implementation of "Increasing diversity of omni-directional images generated from single image using cGAN based on MLPMixer" in ACPR2023.

  • Atsuya Nakata, Ryuto Miyazaki, and Takao Yamanaka, "Increasing diversity of omni-directional images generated from single image using cGAN based on MLPMixer," ACPR2023.

GeneratorImage_v3.png

Sample Images

generated_image_sample.png

Model architecture

model_architecture_v9.png

Requirement

  • Python 3.8 or avobe
  • PyTorch 1.10.2 or avobe

Installation

To install the required packages, run the following command:

pip install -r requirements.txt

Additionally, make sure to install PyTorch from https://pytorch.org/get-started/locally/.

If you are using CUDA 11.3, you can install PyTorch with the following command:

pip3 install torch==1.10.2+cu113 torchvision==0.11.3+cu113 torchaudio==0.10.2+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

Dataset

The data can be downloaded from the following url : Google Drive

Usage

To train the model, use the following command:

python train.py

Evaluation

To evaluate the trained generator model, follow these steps:

  1. Download model weight and calcurated statistics of sun360 from this link : Google Drive.
  2. Copy the trained generator to the eval/generators directory and rename it to "test.pth".
  3. Run the following command from the root directory:
cd ./eval
python calc_all_metrics.py

The evaluation are stored in "eval/evaluation" folder.

References

About

The official pytorch implementation of "Increasing diversity of omni-directional images generated from single image using cGAN based on MLPMixer" in ACPR2023

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages