Changing model size does not always result in the correct model being selected

[ChrisMaunder]

To be investigated.

[melyux]

On 2.8.0, changing the model itself also wouldn't result in the correct model being selected (MobileNet → YOLO, for example)

[melyux]

Any update on this?

[mpolulich]

@seth-planet @mailseth (Sorry, don't know which one you monitor)
I noticed you had some Pull requests and many changes. Was this one of the items fixed?

Hope you don't mind I have a few more questions.

I also saw in IPCamTalk you mentioned that you are running these tests with the objectdetection_coral_mulitpu.py file on the CLI, independent of most of the rest of CPAI. Can you explain how you are doing that so I can try myself? For example, the call with parameters. The comments in the code look like they are just for a non-segmented model. I'm trying to run this with multiple TPUs (at least 2 for now maybe more in future).

Were you ever able to get a model with just security camera classes? Possibly with license plates/vehicle registrations? I actually sent you a DM in IPCamTalk a LONG time ago. Do you ever go back there?

Thanks for everything you have done.