Bug with Max Predictions in AutoML Batch Predictions (image object detection)

[MakeMaday]

I am not able to get the batch prediction API for an AutoML image object detection request to recognize a "maxPrediction" parameter. I am running the prediction against an image that contains several hundred objects.

The response I am getting back is showing 100 object detections regardless of what I specify as the  "maxPredictions" parameter.

I am using the exact "curl" approach specified here, where the "maxPredictions" parameter is mentioned in the API method:
https://cloud.google.com/vertex-ai/docs/image-data/classification/get-predictions?authuser=0#api_2

Background

The documentation here specifies that the default number of bounding boxes returned from a prediction request is 100, and the maximum is 500:
https://cloud.google.com/vertex-ai/docs/image-data/object-detection/prepare-data?authuser=0

The ImageObjectDetection definition shows support for a "maxPredictions" parameter:
gs://google-cloud-aiplatform/schema/predict/params/image_object_detection_1.0.0.yaml

[MakeMaday]

I tested this issue with the online prediction approach and ran into a similar limitation. Using the Python API I can get an accurate response as long as the `max_predictions` parameter is set to a value of 100 or less. Specifying a `max_predictions` parameter with a value greater than 100 returns a maximum of 100 object bounding boxes.