Image Picker upgrade

[neomorphic]

Enhancements requested by Steve

1. Allow user to input a REST api
2. Allow user to right click or otherwise choose a point
3. That point (and the data source) should be posted to the REST API as JSON (choose some format and I will conform)
4. The return value will either be one of the following:
   a) a new data source and x,y,z point
   b) a list of x,y,z points with jpeg thumbnails (embed in return json?)
   If a), pop-up neuroglancer in the same workspace (but maybe 1/4 the size of the original neuroglancer and just an XY view by default) if b) show a list of x,y,z with thumbnails

[stephenplaza]

Revised specs:

inputs: location of top level cloud function (what we have been using) and dataset

• when a user clicks in neuroglancer grayscale with a given modifier or button pressed, call a cloud function which will send the x,y,z and dataset name to the given endpoint
• the response will be a list of x,y,z results (with confidence scores) and/or a message ("no results found", "no signature data for the chosen coordinate", "no good results found"
• provide a paginated list of matches (the function might return more matches than you want to show on a page, maybe only show 10 at a time)
• for each match, show the x,y,z (clicking will navigate to that spot of neuroglancer) and a confidence bar that shows how good the match is
• for each match download a thumbnail from the appropriate api in the cloud function

[stephenplaza]

image picker part 2:

• add a radio button (or something like it) to choose a mode of picking
• mode 1: "query by example" and mode 2: "apply transfer network"
• mode 1 was discussed above
• mode 2 will simply call the appropriate endpoint with the given x,y,z and dataset
• mode 2 could take a while to run (run in background?); the function will return a web address which can then be opened as a seperate window or clicked on by the user to pop up another window (this will just be a link for another neuroglancer outside of the hub). So maybe you can think of executing this as a 'background job' that will show up in a 'job queue'. This stuff should "infinitely" scale on the back-end so allowing multiple jobs is fine and just add results to the result box.
• mode 2 could also return a result of 'not supported'

[stephenplaza]

I can probably add an endpoint that will let you know if the mode are 'available' for a given dataset. Because this might allow us to 'disable' this plugin if there is not support for a dataset.

[neomorphic]

@stephenplaza Could I have more description around this point:

for each match, show the x,y,z (clicking will navigate to that spot of neuroglancer) and a confidence bar that shows how good the match is

Specifically what the confidence bar should be showing? I am guessing the score value. Also where should this bar be located, with the thumbnail?

[stephenplaza]

My thought was something like the old-style search result score shown next to results from search engines. A very low-resolution match score bar next to each icon. The score value should be high for a good match and low for a bad. The highest score is 1.0.