Support all detector modes #95
Conversation
| for i, roi in enumerate(rois): | ||
| geometry = rois[i]["geometry"] | ||
| # TODO add validation to calculate x and y automatically | ||
| x = 0.5 * (geometry["left"] + geometry["right"]) |
There was a problem hiding this comment.
Why are we calculating the centers here?
There was a problem hiding this comment.
The ImageQuery model from the SDK requires x and y to be specified. Tyler thinks there's a way to get pydantic to do this automatically, but for now this seems easiest.
There was a problem hiding this comment.
Yeah apparently Avi asked for the sdk to return the centers of the bboxes as well as left, top, right, bottom coordinates. Tim likes this too.
|
|
||
| if multi_predictions is not None and predictions is not None: | ||
| raise ValueError("Got result with both multi_predictions and predictions.") | ||
| if multi_predictions is not None: |
There was a problem hiding this comment.
That's the field where multiclass/counting results are returned through.
| image_query = create_iqe( | ||
| detector_id=detector_id, | ||
| result_type=result_type, | ||
| label=results["label"], |
There was a problem hiding this comment.
This is going to change soon, you'll need to populate this field according to the mode type
| @@ -25,15 +28,15 @@ def create_iqe( | |||
| created_at=datetime.utcnow(), | |||
| query=query, | |||
| detector_id=detector_id, | |||
| result_type=ResultTypeEnum.binary_classification, | |||
| result_type=result_type, | |||
| result=BinaryClassificationResult( | |||
There was a problem hiding this comment.
We'll need to use a different type here for counting modes
Add support to the edge endpoint for all detector modes. Adds some basic tests to ensure responses are being parsed correctly.