For now, the rule-based binder algorithm seems to be doing a decent job at least for the majority of video we've used (although, we have to wait to see some numbers from evaluation next week).
But for the future, and for generalization of the app beyond some AAPB video samples, we might need ML-based component (or a new e2e model). This issue is a reminder for the future work.
For now, the rule-based binder algorithm seems to be doing a decent job at least for the majority of video we've used (although, we have to wait to see some numbers from evaluation next week).
But for the future, and for generalization of the app beyond some AAPB video samples, we might need ML-based component (or a new e2e model). This issue is a reminder for the future work.