Online Multimodal Fusion

[celikbasak]

Overview

Online Multimodal Fusion for BciPy. The integration and testing of matrix copy phrase task.
This PR is ready for review.

Ticket

https://www.pivotaltracker.com/story/show/186593914

Contributions

• Gaze Evidence Evaluator is generated
• Cleanup & update of gaze signal models
• Cleanup & update of offline analysis
• Copy phrase task is updated to include gaze evidence
• Task evaluator is tested for multimodal fusion

Test

Online tests complete for Multimodal Copy Phrase task

Documentation

Readme updated for gaze models

Changelog

CHANGELOG.md will be updated after PR review

[tab-cmd]

This PR doesn't seem to address any of the issues. Perhaps some commits were lost? I would branch off 2.0.0 and ensure at a minimum the following changes are present:

1. We need the eye tracking model to save as a pkl in the event a user accepts the fusion analysis. It does not currently save - it only prints the 'results'.
2. We need to test whether the existing code can be used to utilize both pkl files when both devices are connected. This should not need any changes. If it does, we should discuss fixes together. If no issues, please report on success/ review the exported data with us.

Let me know the estimated timeline for this. If it's longer than a week, we should roll back the eye tracking code to release 2.0 on schedule. Thanks!

[celikbasak]

This PR doesn't seem to address any of the issues. Perhaps some commits were lost? I would branch off 2.0.0 and ensure at a minimum the following changes are present:

1. We need the eye tracking model to save as a pkl in the event a user accepts the fusion analysis. It does not currently save - it only prints the 'results'.
2. We need to test whether the existing code can be used to utilize both pkl files when both devices are connected. This should not need any changes. If it does, we should discuss fixes together. If no issues, please report on success/ review the exported data with us.

Let me know the estimated timeline for this. If it's longer than a week, we should roll back the eye tracking code to release 2.0 on schedule. Thanks!

I also realized after branching off 2.0.0 some of my work is lost, I will make sure the offline analysis is refactored and the model is saved as a pkl file. I can do another copy phrase test tomorrow with the equipment, but if there are errors we might need to roll back the eye tracking code as you mentioned, as I will need to coordinate a bugfix session with you after tomorrow.