About
Diorama is a framework that allows quick prototyping of real-time or offline computer vision/graphics algorithms.
Typically, prototyping new algorithms requires writing lots of code that are only orthogonally related to the algorithm. For example, real-time 3D human pose tracking from depth sensors (e.g.: Kinect) typically has many goals: (1) run the core algorithm as fast as possible, (2) render the tracked results using OpenGL or some other graphics library, (3) Additionally, one may want to load configuration files, have a GUI for adjusting parameters on the fly, run in headless mode (i.e. no GUI) etc.
In my experience (2) and (3) take a disproportionately large amount of time. This is frustrating because, as a researcher, I find (1) is the most exciting part that I want to spend my time on.
Diorama aims to make (2) and (3) less time consuming by providing a framework that can be reused for different projects and that automates many of the mundane GUI, OpenGL, config file loading, etc. functionality.
At this stage, the expectation is that Diorama will follow some variant of the Model-View-Controller/Presenter software design pattern. Diorama exposes just a methods for you to get started with your project.
Draw() - A separate thread that draws to a OpenGL window
Step() - A separate thread that does the core processing (of algorithm)
Dependencies
C++11 is used extensively, so please use a compiler that supports all of its features. It also depends on:
- cmake (2.8+)
- OpenGL
- Qt (4.8.0+, probably not Qt 5)
Installation
Coming soon...
Contact
Srinath Sridhar (srinaths@umich.edu) Max Planck Institute for Informatics