tentone / depthset

Depth image dataset generator from 3D models for machine learning experiments.

Depthset

• Depth image dataset generator from 3D models.
• Generates big datasets for machine learning research using obj and stl files.

Usage

• Clone the repo.
• Install dependencies by running 'npm install'.
• Copy obj and stl files to the data folder.
• Run the program by calling 'npm run start'.
• Results will be written to the output folder.

Configuration

• In the main.js file its possible to configure the following parameters
  • input = Input folder
  • output = Output folder
  • resolution = Resolution of the images (256 px by default)
  • repetions = How many repetitions with different rotation for each model (10 by default)
  • drawImages = If the images should appear on the window for preview (false by default)
  • near = Camera near plane (0.3 by default)
  • far = Camera far plane (2.0 by default)
  • fov = Camera field of view (60.0 by default)
  • distance = Camera distance to the model (1.2 by default)

Example

• Example of some data generated by this program using screw 3D models.

Dependencies

• NWJS
• three.js

License

• Code is distributed under MIT license (available on GitHub).