Improvement for social experiment visualiser

SainsburyWellcomeCentre / aeon_experiments

Experiment workflows for Project Aeon

BSD 3-Clause "New" or "Revised" License

2 stars 0 forks source link

Improvement for social experiment visualiser #391

Closed Dario55 closed 9 months ago

Dario55 commented 11 months ago

By running the social experiment I realised that there are two things that if added may improve readability on mouse progression and performance online:

[x] Keep track of how many blocks have been done up to that moment
[x] #455
- In the textbox for the current block this would say e.g.: "patch 1 (easy): patch 2 (medium) patch 3 (hard)" for blocks of different thresholds or e.g. "patch 1 (easy) patch 2 (easy) patch 3 (easy)" for blocks of the same threshold
[x] #449
[x] #450
[x] #454

Dario55 commented 11 months ago

proposalvisualisersocial.pptx

Dario55 commented 11 months ago

Additionally, I would propose to change the sleap visualiser from plotting the raw output of the model to plotting the best prediction for each class. If one class is not detected in a given frame, we could apply the heuristic that then the id of the undetected mouse is the one of mouse the mouse not detected.

jkbhagatio commented 10 months ago

For now we'll just post info about the current block in the workflow GUI, and we'll put "total epoch" info in the daily plots, and can discuss how to add it to workflow GUI in future.

jkbhagatio commented 9 months ago

For now we'll just post info about the current block in the workflow GUI, and we'll put "total epoch" info in the daily plots, and can discuss how to add it to workflow GUI in future.

This has been done in #442. We can reopen if we want to discuss