Demonstrations

Deep learning frameworks

  1. DeepLearningTools, a high level framework based on Tensorflow to prototype, train, and deploy deep learning models, enforcing reproducible experiments and systematic good practices that makes models more confident. My tool to explore a variety of problems.

  2. Gammalearn, published here, a high level framework based on PyTorch (Lightning) with similar ideas as DeepLearningTools but more dedicated to Gamma rays analysis for Astrophysics. Born with a collaboration with the LAPP lab and Mikael Jacquemont phD, very nice work !

Image processing retina model, maintained version on OpenCV

  • Retina model is now available under BSD-like licensing in OpenCV !

Please regularly check out git version of OpenCV

Checkout the documentation in the contrib module and the related tutorial

http://docs.opencv.org/

Many demos are provided :

    • Retina processing for image and videos see :

opencv/samples/cpp/retinademo.cpp

  • High Dynamic Range images compression (tone mapping) see sample :

opencv/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping

  • High Dynamic Range images sequence compression (tone mapping) see sample :

opencv/samples/cpp/OpenEXRimages_HighDynamicRange_Retina_toneMapping_video

Older retina versions

(no more maintained, based on OpenCV1.0)

  • GUI version online but not maintained (prefer the OpenCV version), it includes :

  • _V1 cortex demo and motion event detection added !

WARNING, current demo version presents HDR bugs (output is not scaled appropriately, code is corrected, but i am waiting for a more compatible OpenCV 2.0 library to make it available)

  • User friendly Retina Demo:

The retina model is developped in C++ in a multi-threaded application. This version allows retina parameters to be modified freely. A Qt 4.5 interface has been developped for easy input load and model parameters adjustment. You can load images, image sequences, videos or use a connected webcam.

You can experiment retina model image processing: extract the detail and motion channels and also perform Image Tone Mapping (use the tone mampping button in the general frame to get generic tone mapping parameters).

Load, adjust parameters and run! Parameters can be adjusted while a process is running. Experiment the way you want! Some tooltips are added to help you understand the role of each parameter, if it is not clear, contact me.

Installation issues: you should not have other stuff to install. Try to install vcredist-x86 libraries from Microsoft if the program does not launch ONLY. Contact me for any other inquiries.

The Windows XP/Vista version is here, version 1.0.

Note: the retina model can be used for tone mapping processing. See publication pages for citation/explanations.

Screenshots:

General interface

Retina setup (adjust parameters before program start or during processing)