I was able to solve the issue, but I guess there is a mistake in the version on Azure Custom Vision Object Detector Quick Start.
The call below uses the variable "ENDPOINT" which is set to the "VISION_TRAINING_ENDPOINT" which is actually for the trainer. The ENDPOINT for the predictor should be set to "VISION_PREDICTION_ENDPOINT". I was able to solve the issue by this minor change. Please correct this mistake in the Quick Start guide as well.
predictor = CustomVisionPredictionClient(ENDPOINT, prediction_credentials)