Home
Vision scientists have discovered many facts and developed many precise computations that characterize critical steps in vision. Some of these computations are complex and require access to numerical tables (e.g., color-matching, optical wavefront aberrations). ISETBIO provides many computations and the necessary numerical data to perform these calculations. We aim to provide the information in a convenient and extensible format so that investigators can both use and extend vision science. We hope this code helps to develop new biological insights and theoretical ideas.
We further hope that ISETBIO will be a computational framework that helps specialists in different branches of vision science and engineering explain and share their work. Specialists working on one part of the system - say optics - can contribute code that can be used by people who are expert in a different part of the system - say the retina. By placing the knowledge of the vision science fields in a unified software framework, we believe that vision scientists will be able to better see the big picture
This project is based on the idea that computational models that are approximately right and exist are preferable to models that are perfect and do not exist. We hope that our colleagues and students accept that a specific computation that implements our best current understanding is better than no computation. To support this approach, it is important to specify the domain over which the calculations have been checked and to give some sense of the accuracy of the model over that domain. We try to indicate the stimulus compliance range and precision of computations, as best we can, throughout the code and in the documentation. If we identify a problem, we try to follow this advice.
The ISETBIO code base is evolving and being checked by different individuals. We do our best to write validation tests for each of the components as well as system tests and regression tests. We have confidence in many of the critical routines, which have been independently implemented and produce the same results. Other routines are still being tested. We have tried to make the differences clear by comments in the code and on this wiki. We invite you to perform your own tests and to share the results with us.
This project includes specifications of the scene radiance, optics and retinal irradiance, cone mosaic absorptions and photocurrent, retinal encoding, eye movements, and computational classifiers. The compliance range and precision of the scene and optics and absorptions are the most advanced, and our confidence declines as we dive deeper into the nervous system.
Our ambitions go beyond the current coverage. With more experience and feedback, we will decide how much further we can extend the scope of these computations. That's right LGN and V1, we're talking about you.
If you use ISETBio, please cite it using the following recent publication:
- Nicolas P. Cottaris, Haomiao Jiang, Xiaomao Ding, Brian A. Wandell, David H. Brainard; A computational-observer model of spatial contrast sensitivity: Effects of wave-front-based optics, cone-mosaic structure, and inference engine. Journal of Vision 2019;19(4):8. doi: 10.1167/19.4.8.
- Brainard Lab Illumination Discrimination Calculations
- Brainard Lab Hyperspectal Image Computations
- Brainard Figure Code
- Brainard/Stockman Book Figure Code
- Chichilnisky RGC Experiments
- Predictions of Color Thresholds
- Wandell Lab Letter Classification
- Wandell Lab Vernier Acuity
- Winawer Lab Gabor Analyses