We have currently two possible ways of identifying the subject entering the nest, either visual classification using a SLEAP model or an RFID event. Either way, we can use IDs which are triggered before the weight scale stabilizes as handles to associate with the weight readout.
Alternatively, we could try to implement a list of animals in the nest using the following strategy:
Add animal to list when RFID is triggered at the nest entrance:
main worry is false negative at nest entrance; could potentially complement with visual tracking but may be unreliable
in Autonomouse this was solved by having a location where animals have to go spend some time, e.g. RFID at water delivery place to lower probability of false negative at that location
Remove animal from list when visual SLEAP tracking model detects animal "sufficiently far away"
Weight readouts would be taken only when there is exactly one animal in the list
The above strategy could be refined by having more RFID antennas in the corridor to provide extra "exit" events
Potentially we can also compare this weight with previous weights from the same animal to check if we are getting consistent reads. If the weight is not consistent after 3 or 4 reads we could trigger an alert for further investigation.
We have currently two possible ways of identifying the subject entering the nest, either visual classification using a SLEAP model or an RFID event. Either way, we can use IDs which are triggered before the weight scale stabilizes as handles to associate with the weight readout.
Alternatively, we could try to implement a list of animals in the nest using the following strategy:
Potentially we can also compare this weight with previous weights from the same animal to check if we are getting consistent reads. If the weight is not consistent after 3 or 4 reads we could trigger an alert for further investigation.