MaxRoiPool
Description
The ROI max pool consumes an input tensorX and a region of interest (ROI) to apply the max pool on each ROI to produce the output 4-D shape tensor (num_roi, channels, pooled_shape[0], pooled_shape[1]).
Input
Two inputs:
x: tensor of type float16.
rois: tensor of type float16.
Output
One output:
y: tensor of type float16.
Property
pooled_shape: list of ints.
spatial_scale: float. Defaults to 1.0.
Constraints
The atc tool does not support the input of the float type when the --precision_mode=must_keep_origin_dtype parameter is used.
ONNX Opset Support
Opset v8/v9/v10/v11/v12/v13/v14/v15/v16
Parent topic: Operators Specifications of the AI Framework