When I say marked-up or burned, I'm talking about this setting:
My use case is training my own model. I'd still like push notifications and email to come through with marked up images so I can quickly tell what the object is, where it was detected, if it's a false positive, etc. However, I'd also like to use that image to reinforce a new model, which won't work if there's a giant confirmation box burned into the image. If I have markup enabled, it applies to all on alert actions though, such as uploading the image to FTP or in the alerts directory. I'm wondering if there's a workaround like using a Do Command action or something.
So far I've tried:
- Alert actions -> FTP Upload of the camera image (not the alert image, since this has the burn in). This tends to be a few seconds behind the actual trigger image though, and often misses the action.
- Alert -> Add to the alerts list, "Hi-res JPEG files". This doesn't seem to do anything different from add to database only. I only see the original alert image with the markup burn in the Alerts folder, and the image that's created looks like the same resolution as the standard images.
- Record -> JPEGs isn't ideal. Using "When alerted" will often lag a few seconds behind, kind of like that of the FTP upload of the camera image when the alert triggers. By the time the trigger fires, the object may be gone. "When triggered" doesn't work for me, both because the timing is off and because that would create a ton of false positive images.
