Real Synthetic Dataset Generation Tool

Real Synthetic Dataset Generation Tool

1. Introduction

This tool is used to generate synthetic datasets for the purpose of testing and benchmarking machine learning algorithms. The tool is written in Python and uses the OpenCV library for data manipulation and NumPy for numerical operations. The tool is designed to be used in a Python environment.

1.1. Workflow of the tool

Here is the self-explanatory workflow of the tool.

2. Installation

2.0. Pre-requisites

Prepare the data and extract Transparent Image Mask from this repository Transparent Image Extraction using Mask (Pipeline-1)

2.1. Requirements

The tool requires Python 3.7 or higher.

2.2. Required Libraries

pip install requirements.txt

2.3. Folder Hierarchy

├── Dataset
│
│   ├── background_images
│   │   ├── 1.jpg
│   │   ├── 2.jpg
│   │   ├── 3.jpg
│   
│   ├── input
│   │   ├── 0-class-0 [folder name starting with class number]
│   │   │   ├── 1.png
│   │   │   ├── 2.png
│   │   │   ├── 3.png
│   │   ├── 1-class-1
│   │   │   ├── 1.png
│   │   │   ├── 2.png
│   │   │   ├── 3.png
│   │   ├── 2-class-2
│   │   │   ├── 1.png
│   │   │   ├── 2.png
│   │   │   ├── 3.png
│   
│   ├── Dataset_output
│   │   ├── 1.jpg
│   │   ├── 1.txt
│   │   ├── 2.jpg
│   │   ├── 2.txt
│   │   ├── 3.jpg
│   │   ├── 3.txt

3. Usage

The tool is designed to be used in a Python environment.

python setup.py

3.1 Object Selection GUI

3.2 Object Selected

3.3 Click on Base Image

3.4 Object Placed

3.5 Auto Generated Label files in YOLO format

4. License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.