ImageBgRemoval
Overview
This project is to detect the several objects including person in an image, convert white background to transparent and save the obtained image into png file format. To improve the accuracy of transformation, the Pytorch framework is used, which needs usage of GPU. In detail, the deeplab framework is used in this project.
Structure
-
data
There are two directories, one of which contains the images to convert background and other of which shows the result images, contains the png file.
-
source
The main source code to process image transformation.
-
utils
The tools concerned with image processing, models and file management.
-
main
The main execution file.
-
requirements
All the dependencies to execute project.
-
settings
Several options in it.
Installation
-
Environment
Ubuntu 18.04, Python 3.6, GPU
-
Dependency installation
pip3 install -r requirements.txt -
Preparation for deeplab model
- Please download the model from https://drive.google.com/file/d/1vzLRbMua_hxYXpBH8WWTozThHX4DTvIX/view?usp=sharing and copy them to /utils/model.
- After downloading deep learning model at http://download.tensorflow.org/models/deeplabv3_pascal_train_aug_2018_01_04.tar.gz, please copy it to utils/model directory.
Execution
-
Please copy images to transform in data/input directory.
-
Please go ahead to this project directory and run the following command in terminal
python3 main.py
Then this project transforms all the images in input directory and exports the result with png format.