Skip to content

code for paper"A High-precision Semantic Segmentation Method Combining Adversarial Learning and Attention Mechanism"

Notifications You must be signed in to change notification settings

Tong-Gao-dream/UAGAN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

PyTorch implementation of UAGAN(U-net Attention Generative Adversarial Networks)

This repository contains the source code for the paper "A High-precision Semantic Segmentation Method Combining Adversarial Learning and Attention Mechanism"

Author email: [email protected]

Github Github

Datasets

We use the renderings of ShapeNet and Carvana in our experiments,which are available below:

Model Structure

Prerequisites

Clone the Code Repository

git clone https://github.com/clfs0003/UAGAN

Install Python Denpendencies

cd UAGAN
pip install -r UAGAN_requirements.txt

Update Settings in config.py

You need to update hyperparametersthe of the model and path of the datasets :

    parser.add_argument('--image_dir', type=str, default='./dataset/img', help='input RGB image path')
    parser.add_argument('--mask_dir', type=str, default='./dataset/mask', help='input mask path')
    parser.add_argument('--lr', type=float, default='0.0002', help='learning rate')
    parser.add_argument('--batch_size', type=int, default='5', help='batch_size in training')
    parser.add_argument('--b1', type=float, default=0.5, help='adam: decay of first order momentum of gradient')
    parser.add_argument("--b2", type=float, default=0.999, help="adam: decay of first order momentum of gradient")
    parser.add_argument("--epoch", type=int, default=600, help="epoch in training")

Part of the experimental results are as follows

results of ShapeNet: results of ShapeNet results of Carvana: results of Carvana

Get Started

To train UAGAN, you can simply use the following command:

python3 main.py

Note: Since our paper has not been published yet, we only show the model structure and training code. When the paper is published, we will publish the full work of the paper. Welcome scholars to discuss and exchange.

License

This project is open sourced under MIT license.

About

code for paper"A High-precision Semantic Segmentation Method Combining Adversarial Learning and Attention Mechanism"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages