A question for a proof of concept i am working on, would i in any way be able to add Object Detection and Anomaly Detection blocks to the same impulse? If so, how should i set up my data in the studio? I’m thinking that Anomaly Detection would be useful to help identify false positives where one would be catastrophic, would this be viable or would it be overkill?
Given the two different labelling methods required for visual anomaly detection and object detection, my recommendation would be to create two separate projects and models for each of these problems, with the same dataset but labeled differently.
When deploying if you’re using a Linux device then running both models at the same time is easy with something like our Linux Python SDK https://github.com/edgeimpulse/linux-sdk-python/tree/master, simply initialise two ImpulseRunner objects with each of your models.
If you’re deploying to MCU things are a bit more complex but still possible https://docs.edgeimpulse.com/tutorials/topics/inference/run-multiple-impulses-cpp