ScopeFastrcnnPredictionsPass

Description

Fuses the fastrcnn_predictions Scope operator into a FastrcnnPredictions operator. The scope includes a multiple of two tf.TopKV2 operators, a multiple of three tf.Where operators, tf. NonMaxSuppressionV2 operators, tf.Less operators, and tf.LoopCond operators. The ExpandDims, clip_boxes, and decode_bbox_target operators are excluded.

Scope Details

Result Operator Prototype

FastrcnnPredictions. For details, see Operator Acceleration Library API Reference.

Fusion Mapping

The inputs of fastrcnn_predictions/transpose and fastrcnn_predictions/GatherNd are used as the input rois after fusion.

The input of fastrcnn_predictions/strided_slice is used as the input score after fusion.

The output of fastrcnn_predictions/TopKV2 is used as the output sorted_rois after fusion.

The output of fastrcnn_predictions/GatherNd is used as the output sorted_scores after fusion.

The output of fastrcnn_predictions/Add is used as the output sorted_classes after fusion.

The output of fastrcnn_predictions/strided_slice is used as the input of fastrcnn_predictions/transpose_1.

The output of fastrcnn_predictions/transpose_1 is used as the inputs of fastrcnn_predictions/map and fastrcnn_predictions/boolean_mask.

The output of astrcnn_predictions/map is used as the input of fastrcnn_predictions/Where.

The output of fastrcnn_predictions/Where is used as the input of fastrcnn_predictions/Gather.

The output of fastrcnn_predictions/boolean_mask is used as the inputs of fastrcnn_predictions/Size and fastrcnn_predictions/TopKV2.

The output of fastrcnn_predictions/Size is used as the input of fastrcnn_predictions/Minimum.

The output of fastrcnn_predictions/Minimum is used as the input of fastrcnn_predictions/TopKV2.

fastrcnn_predictions/TopKV2 is used as the output sorted_rois after fusion and input of fastrcnn_predictions/Gather.

The output of fastrcnn_predictions/Gather is used as the input of fastrcnn_predictions/filtered_indices.

The output of fastrcnn_predictions/filtered_indices is used as the inputs of fastrcnn_predictions/GatherNd and fastrcnn_predictions/ToFloat.

The output of fastrcnn_predictions/ToFloat is used as the input of fastrcnn_predictions/strided_slice_1.

The output of fastrcnn_predictions/strided_slice_1 is used as the input of fastrcnn_predictions/Add.

Applicable Network

2D-H1

Fusion Pattern Type

Non-general fusion pattern