Reputation: 1
I already tested all the depthai demos listed on https://docs.luxonis.com/en/latest/#demo-script) on my OAK device. Every demo is working fine but when I’m using the Luxonis OAK code provided by Roboflow (Luxonis OAK (On Device) - Roboflow) for my object detection project on my Raspberry Pi 4 I always get the Key-Error ‘colors’, more detailed:
Traceback (most recent call last): File “home/pi/myvideo.py”, line 8, in rf=RoboflowOak(model=“schwimmen-cp33a”, confidence=0.05, overlap=0.5, File “/usr/local/lib/python3.9/dist-packages/roboflowak/init.py”, line 46, in init self.colors=self.model_objects[“colors”] KeyError: ‘colors’
Roboflowoak is on the latest version 0.0.8, python at 3.9.
I’m looking foreward for your suggestions.
Best regards, Gerald
already tested all the depthai demos listed on https://docs.luxonis.com/en/latest/#demo-script already updated Roboflowoak
Upvotes: -2
Views: 107
Reputation: 1
This bug has been fixed on the Roboflow backend. Your current setup should now work as expected!
Updated for clarity:
Roboflowoak version >0.0.8 will work as expected, regardless of whether the underlying Roboflow dataset has ‘colors’ prop on the model objects. The api call has been updated to always return a map of classnames to colors in the ‘colors’ on model_objects
. Future issues can be filed directly on Roboflow's bugtracker: https://github.com/roboflow/external-bugtracker/issues.
Upvotes: 0