Compute Library
 20.08
NEDetectionPostProcessLayer.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2019-2020 Arm Limited.
3  *
4  * SPDX-License-Identifier: MIT
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to
8  * deal in the Software without restriction, including without limitation the
9  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10  * sell copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in all
14  * copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
25 
26 #include "arm_compute/core/Error.h"
29 
30 #include <cstddef>
31 #include <ios>
32 #include <list>
33 
34 namespace arm_compute
35 {
36 NEDetectionPostProcessLayer::NEDetectionPostProcessLayer(std::shared_ptr<IMemoryManager> memory_manager)
37  : _memory_group(std::move(memory_manager)), _dequantize(), _detection_post_process(), _decoded_scores(), _run_dequantize(false)
38 {
39 }
40 
41 void NEDetectionPostProcessLayer::configure(const ITensor *input_box_encoding, const ITensor *input_scores, const ITensor *input_anchors,
42  ITensor *output_boxes, ITensor *output_classes, ITensor *output_scores, ITensor *num_detection, DetectionPostProcessLayerInfo info)
43 {
44  ARM_COMPUTE_ERROR_ON_NULLPTR(input_box_encoding, input_scores, input_anchors, output_boxes, output_classes, output_scores);
45  ARM_COMPUTE_ERROR_THROW_ON(NEDetectionPostProcessLayer::validate(input_box_encoding->info(), input_scores->info(), input_anchors->info(), output_boxes->info(), output_classes->info(),
46  output_scores->info(),
47  num_detection->info(), info));
48 
49  const ITensor *input_scores_to_use = input_scores;
50  DetectionPostProcessLayerInfo info_to_use = info;
51  _run_dequantize = is_data_type_quantized(input_box_encoding->info()->data_type());
52 
53  if(_run_dequantize)
54  {
55  _memory_group.manage(&_decoded_scores);
56 
57  _dequantize.configure(input_scores, &_decoded_scores);
58 
59  input_scores_to_use = &_decoded_scores;
60 
61  // Create a new info struct to avoid dequantizing in the CPP layer
62  std::array<float, 4> scales_values{ info.scale_value_y(), info.scale_value_x(), info.scale_value_h(), info.scale_value_w() };
63  DetectionPostProcessLayerInfo info_quantized(info.max_detections(), info.max_classes_per_detection(), info.nms_score_threshold(), info.iou_threshold(), info.num_classes(),
64  scales_values, info.use_regular_nms(), info.detection_per_class(), false);
65  info_to_use = info_quantized;
66  }
67 
68  _detection_post_process.configure(input_box_encoding, input_scores_to_use, input_anchors, output_boxes, output_classes, output_scores, num_detection, info_to_use);
69  _decoded_scores.allocator()->allocate();
70 }
71 
72 Status NEDetectionPostProcessLayer::validate(const ITensorInfo *input_box_encoding, const ITensorInfo *input_scores, const ITensorInfo *input_anchors,
73  ITensorInfo *output_boxes, ITensorInfo *output_classes, ITensorInfo *output_scores, ITensorInfo *num_detection, DetectionPostProcessLayerInfo info)
74 {
75  bool run_dequantize = is_data_type_quantized(input_box_encoding->data_type());
76  if(run_dequantize)
77  {
78  TensorInfo decoded_classes_info = input_scores->clone()->set_is_resizable(true).set_data_type(DataType::F32);
79  ARM_COMPUTE_RETURN_ON_ERROR(NEDequantizationLayer::validate(input_scores, &decoded_classes_info));
80  }
81  ARM_COMPUTE_RETURN_ON_ERROR(CPPDetectionPostProcessLayer::validate(input_box_encoding, input_scores, input_anchors, output_boxes, output_classes, output_scores, num_detection, info));
82 
83  return Status{};
84 }
85 
87 {
88  MemoryGroupResourceScope scope_mg(_memory_group);
89 
90  // Decode scores if necessary
91  if(_run_dequantize)
92  {
93  _dequantize.run();
94  }
95  _detection_post_process.run();
96 }
97 } // namespace arm_compute
bool is_data_type_quantized(DataType dt)
Check if a given data type is of quantized type.
Definition: Utils.h:1121
void run() override final
Run the kernels contained in the function.
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
virtual DataType data_type() const =0
Data type used for each element of the tensor.
static Status validate(const ITensorInfo *input_box_encoding, const ITensorInfo *input_class_score, const ITensorInfo *input_anchors, ITensorInfo *output_boxes, ITensorInfo *output_classes, ITensorInfo *output_scores, ITensorInfo *num_detection, DetectionPostProcessLayerInfo info=DetectionPostProcessLayerInfo())
Static function to check if given info will lead to a valid configuration of CPPDetectionPostProcessL...
void configure(const ITensor *input, ITensor *output)
Configure the kernel.
1 channel, 1 F32 per channel
Store the tensor's metadata.
Definition: ITensorInfo.h:40
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Status class.
Definition: Error.h:52
Interface for NEON tensor.
Definition: ITensor.h:36
Copyright (c) 2017-2020 Arm Limited.
TensorAllocator * allocator()
Return a pointer to the tensor's allocator.
Definition: Tensor.cpp:48
void manage(IMemoryManageable *obj) override
Sets a object to be managed by the given memory group.
Definition: MemoryGroup.h:79
static Status validate(const ITensorInfo *input, const ITensorInfo *output)
Static function to check if given info will lead to a valid configuration of NEDequantizationLayer.
void run() override
Run the kernels contained in the function.
void allocate() override
Allocate size specified by TensorInfo of CPU memory.
virtual std::unique_ptr< T > clone() const =0
Provide a clone of the current object of class T.
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor's metadata.
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
Memory group resources scope handling class.
Definition: IMemoryGroup.h:82
void configure(const ITensor *input_box_encoding, const ITensor *input_score, const ITensor *input_anchors, ITensor *output_boxes, ITensor *output_classes, ITensor *output_scores, ITensor *num_detection, DetectionPostProcessLayerInfo info=DetectionPostProcessLayerInfo())
Configure the detection output layer CPP function.
void configure(const ITensor *input_box_encoding, const ITensor *input_score, const ITensor *input_anchors, ITensor *output_boxes, ITensor *output_classes, ITensor *output_scores, ITensor *num_detection, DetectionPostProcessLayerInfo info=DetectionPostProcessLayerInfo())
Configure the detection output layer NE function.
NEDetectionPostProcessLayer(std::shared_ptr< IMemoryManager > memory_manager=nullptr)
Constructor.
Detection Output layer info.
Definition: Types.h:1062
Store the tensor's metadata.
Definition: TensorInfo.h:45
void run() override
Run the kernels contained in the function.
static Status validate(const ITensorInfo *input_box_encoding, const ITensorInfo *input_class_score, const ITensorInfo *input_anchors, ITensorInfo *output_boxes, ITensorInfo *output_classes, ITensorInfo *output_scores, ITensorInfo *num_detection, DetectionPostProcessLayerInfo info=DetectionPostProcessLayerInfo())
Static function to check if given info will lead to a valid configuration of NEDetectionPostProcessLa...