hl123-123 / SMIS

Semantically Multi-modal Image Synthesis(CVPR 2020)

Semantically Multi-modal Image Synthesis

Project page / Paper / Demo

Semantically Multi-modal Image Synthesis(CVPR2020).
Zhen Zhu, Zhiliang Xu, Ansheng You, Xiang Bai

Requirements

---

• torch>=1.0.0
• torchvision
• dominate
• dill
• scikit-image
• tqdm
• opencv-python

Getting Started

---

Data Preperation

DeepFashion
Note: We provide an example of the DeepFashion dataset. That is slightly different from the DeepFashion used in our paper due to the impact of the COVID-19.

Cityscapes
The Cityscapes dataset can be downloaded at here

ADE20K
The ADE20K dataset can be downloaded at here

Test/Train the models

Download the tar of the pretrained models from the Google Drive Folder. Save it in checkpoints/ and unzip it. There are deepfashion.sh, cityscapes.sh and ade20k.sh in the scripts folder. Change the parameters like --dataroot and so on, then comment or uncomment some code to test/train model. And you can specify the --test_mask for SMIS test.

Acknowledgments

---

Our code is based on the popular SPADE