Home

Table of Contents

• Getting Started
  • Prerequisites
  • Installation
• Usage
  • Structure
  • How to use
• Marketing Plan
• Troubleshoots

This wiki is about using repo. If you want to know about the overall project, please refer to this link.

Getting Started

This is an example of how you may give instructions on setting up your project locally. To get a local copy up and running follow these simple example steps.

Prerequisites

This is an example of how to list things you need to use the software and how to install them.

• OpenCV with contrib modules

pip install opencv-contrib-python

• Scipy 1.2.0(Important)

pip install scipy==1.2.0

• Pillow

pip install pillow

• Tensorflow

pip install tensorflow==2.2.0

• Keras

pip install Keras==2.2.0

• Numpy

pip install numpy

Or you can install the libraries you need to use at once. It is suitable when using a new virtual environment (interpreter).

pip install -r requirements.txt

Installation

1. Complete the prerequisite settings

2. Clone the repo

git clone https://github.com/doljae/convert_image_maker.git

3. Unzip the repo and put the folder in the path to use the library.

4. Please refer to the manual to import and use the library.

5. For detailed usage instructions, see user_guide.pdf

Usage

This section describes the overall structure of the project and how to use it.

Structure

./
|   README.md
|   requirements.txt
|   test.py  --> Usage example
|          
+---image
|   |   image_handle.py  --> Image handler class(include extraction, conversion fucntion)
|   |   readme.txt
|   |   
|   +---clean_back
|   |       apple.png --> sample image for conversion using specific object
|   |       
|   +---crop_images
|   |       crop_0.jpg  --> results of object extraction in input image
|   |       crop_1.jpg
|   |       crop_2.jpg
|   |       crop_X.jpg
|   |       crop_location.txt  --> location of cropped images
|   |       cut_image1.jpg  --> The image from which the detected internal objects are extracted.
|   |       cut_image2.jpg  
|   |       
|   +---result_images
|   |       converted_image.jpg  --> The final converted result
|   |       original_image.jpg  --> input image
|   |       readme.txt
|   |                
+---pix2pix
|   |   data_loader.py
|   |   pix2pix.py  --> conversion using GAN
|   |   
|   +---datasets
|   |   \---tmp
|   |       +---saved
|   |       |       crop_0.jpg  --> Internal object already converted due to GAN model
|   |       |       crop_1.jpg
|   |       |       crop_2.jpg
|   |       |       crop_X.jpg
|   |       |       
|   |       \---test
|   +---keras_contrib
|   |     
\---socket
|   |   receive_convert_send.py --> Server, Receive an image to an external device and send the converted image

How to use

Please refer to test.py for specific usage.

from image.image_handle import convertImageMaker
import os
import glob
import random

files = glob.glob('./image/crop_images/*')
for f in files:
    os.remove(f)

convertTest=convertImageMaker()
convertTest.image_extract()
convertTest.image_save_crop_location()
convertTest.image_convert(convert_type=2)
# convertTest.image_convert(convert_type=random.randint(1,4))

1. Import libraries

from image.image_handle import convertImageMaker
import os
import glob
import random

The python module is sensitive to path setting. Please import the module with the correct path through the functions of the IDE.
os, glob, and random are not essential import modules.

2. Initialize convertImageMaker class

convertTest=convertImageMaker()

This class has 3 functions(image_extract(), image_save_crop_location(), image_convert(convert_type=X))

3. Object detection & extraction

convertTest.image_extract()

This function searches for the object inside the input image, extracts it and saves it.
The extracted results are saved in ./image/crop_images/crop_X.jpg. (See Structure)

4. Save the location of extracted object in input image

convertTest.image_save_crop_location()

This function stores the location of the extracted object.
This value is relative to the original image and is represented by four integer values of (x_po, y_po, width, height).
The results are saved in ./image/crop_images/crop_location.txt. (See Structure)

5. Transforms the extracted object and combines it with the original image to return the result.

convertTest.image_convert(convert_type=2)

The conversion method is determined by the convert_type parameter.
The conversion method according to the parameter value is as follows.

convert_type=1 --> Rotation
convert_type=2 --> Color inversion
convert_type=3 --> Substitution
convert_type=4 --> GAN conversion

The target of all conversion methods is the object image extracted from the original image.

6. Result The converted image is saved in ./image/result_images/converted_image.jpg. (See Structure)

Marketing Plan

Troubleshoots

1. Module version dependency
   This project is module version sensitive. We are currently updating the stable release version, including the latest version of TensorFlow.
   Especially, in case of conversion using GAN model, please check the wiki and error log if it does not work properly.

2. Conversion according to GAN model
   The GAN model converts objects to other similar objects. And, like an AI model, there is a possibility that a similar transformation may be made by learning data, or vice versa.
   
   

Of course, the conversion itself is fine, so there is no problem when using only the converted internal object image.
If other problems occur, please register as an open issues.