Synthetic Data Generation

A python module to generate synthetic images from 3D models, for use in image detection/segmentation tasks.

Usage

1. Clone the repo

   git clone https://github.com/amurudkar/synthetic-data-generation.git
   cd synthetic-data-generation

2. Setup your conda or pip environment

   conda env create -f environment.yml

   or

   Create a virtual environment using this guide (optional), then install packages using:

   python -m pip install -r requirements.txt

3. Run the generate.py script

   python ./generate.py -m <path_to_obj_file> -o <output_dir>

Configuration

• Render settings can be configured using the default_render_config.json as a template.
• Additionally, camera properties can be changed by using camera.json file as a template.
• The script uses the files mentioned above as default but different files can be provided using command line arguments. Run python ./generate.py -h to get a full list of arguments.