Welcome to vodet’s documentation!

Variational Object DETector for time-lapse images.

vodet is a simple library for semi-supervised object detection for anything using pixyz and PyTorch. Since Gaussian-Mixture Variational Auto Encoder (GMVAE) is used in the image classifier, it requires a very small amount (two or three) of annotation images. Though vodet assumes time-lapse images, it may work well with other kinds of images if the sizes of objects do not vary by images.

Contents:

• Installation
  • Dependencies
• Algorithm
  • Object searching
  • Label estimation for each patches
• Usage
  • Image Annotating
  • Setting up data directories
  • Creating an GMVAE instance
  • Generating patch images
  • Preparing Dataloaders
  • Model setting
  • Training
  • Detection
  • Utility functions
  • Acknowledgements

Package Reference:

• vodet.gmvae module
• vodet.detect module
• vodet.distributions module
• vodet.utils module

Indices and tables

• Index

• Module Index

• Search Page