mirror of https://github.com/alibaba/MNN.git
				
				
				
			
		
			
				
	
	
		
			57 lines
		
	
	
		
			2.1 KiB
		
	
	
	
		
			C++
		
	
	
	
			
		
		
	
	
			57 lines
		
	
	
		
			2.1 KiB
		
	
	
	
		
			C++
		
	
	
	
| //
 | |
| //  ShapeDetectionPostProcess.cpp
 | |
| //  MNN
 | |
| //
 | |
| //  Created by MNN on 2019/10/29.
 | |
| //  Copyright © 2018, Alibaba Group Holding Limited
 | |
| 
 | |
| #include "core/SizeComputer.hpp"
 | |
| 
 | |
| namespace MNN {
 | |
| 
 | |
| class DetectionPostProcessSize : public SizeComputer {
 | |
| public:
 | |
|     bool onComputeSize(const MNN::Op *op, const std::vector<Tensor *> &inputs,
 | |
|                        const std::vector<Tensor *> &outputs) const override {
 | |
|         MNN_CHECK(inputs.size() == 3, "DetectionPostProcess should have 3 inputs!");
 | |
|         MNN_CHECK(outputs.size() == 4, "DetectionPostProcess should have 4 outputs!");
 | |
| 
 | |
|         auto postProcess           = op->main_as_DetectionPostProcessParam();
 | |
|         const int numDetectedBoxes = postProcess->maxDetections() * postProcess->maxClassesPerDetection();
 | |
| 
 | |
|         const int bathSize = inputs[0]->batch();
 | |
| 
 | |
|         // Outputs: detection_boxes, detection_scores, detection_classes,
 | |
|         // num_detections
 | |
|         auto detectionBoxes                 = outputs[0];
 | |
|         detectionBoxes->buffer().dimensions = 3;
 | |
|         detectionBoxes->setLength(0, bathSize);
 | |
|         detectionBoxes->setLength(1, numDetectedBoxes);
 | |
|         detectionBoxes->setLength(2, 4);
 | |
|         detectionBoxes->buffer().type = halide_type_of<float>();
 | |
| 
 | |
|         auto detectionClasses                 = outputs[1];
 | |
|         detectionClasses->buffer().dimensions = 2;
 | |
|         detectionClasses->setLength(0, bathSize);
 | |
|         detectionClasses->setLength(1, numDetectedBoxes);
 | |
|         detectionClasses->buffer().type = halide_type_of<float>();
 | |
| 
 | |
|         auto detectionScores                 = outputs[2];
 | |
|         detectionScores->buffer().dimensions = 2;
 | |
|         detectionScores->setLength(0, bathSize);
 | |
|         detectionScores->setLength(1, numDetectedBoxes);
 | |
|         detectionScores->buffer().type = halide_type_of<float>();
 | |
| 
 | |
|         auto numDetections                 = outputs[3];
 | |
|         numDetections->buffer().dimensions = 1;
 | |
|         numDetections->setLength(0, 1);
 | |
|         numDetections->buffer().type = halide_type_of<float>();
 | |
| 
 | |
|         return true;
 | |
|     }
 | |
| };
 | |
| 
 | |
| REGISTER_SHAPE(DetectionPostProcessSize, OpType_DetectionPostProcess);
 | |
| 
 | |
| } // namespace MNN
 |