Pre computed OI Architecture - ISET/isetonline GitHub Wiki

Currently isetonline uses a multi-tier architecture:

< Below notes are from original design for OIs. Needs an update. >

ol_dataPrep.m enumerates desired OI and Sensor files, using isetcam to create sensorImages of each pair
sensor parameters such as shutter time are chosen (plenty of room for experimentation)
single exposure, burst exposures, and bracketed are all computed
there is an option for camera motion simulation as a poor substitute for re-rendering the OI
those files, along with data for the lenses being used, are exported in JSON format
Optionally, the data can also be stored in a mongodb, but currently we don't use it for anything
As an example/sample, generated images are also run through YOLOv4 and annotated for use online
rgb previews are also generated using a simplistic IP

App includes a variety of CoreUI components and a file-driven version of Ag-Grid
The app uses the .json files generated by prepdata.m
It allows the user to scroll through, sort, and filter sensorImages by parameter
The user can also see the effects of bracket and burst capture.
The user can also show/hide the results of YOLOv4 on the various sensor images.
The user can also download the raw OI data, sensor info, or sensorImage data

Using an app built with the Express framework, user access pre-compiled pieces of ISET code to allow users to experiment
The front-end can submit an oi Filename and a sensor object and ask for the backend to compute it
The server invokes a pre-compiled Matlab custom function, oi2sensor and writes out the result
It then (will) return the output filename
Right now, the server runs on a different port from the client. Before production they should be integrated