MLECO-2874: Move NMS out of the OD use_case
* Add ImageUtils
* Move image related code from UseCaseCommonUtils to ImageUtils
* Move NMS related code to ImageUtils
* Delete test specific ImageUtils and use new ImageUtils
Signed-off-by: Richard Burton <richard.burton@arm.com>
Change-Id: Icbf5dd9c6a941b0126ecdf69a0c9d9969f22729f
diff --git a/source/use_case/ad/src/UseCaseHandler.cc b/source/use_case/ad/src/UseCaseHandler.cc
index 420e6d4..853ab08 100644
--- a/source/use_case/ad/src/UseCaseHandler.cc
+++ b/source/use_case/ad/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -22,6 +22,7 @@
#include "hal.h"
#include "AdMelSpectrogram.hpp"
#include "AudioUtils.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "AdPostProcessing.hpp"
#include "log_macros.h"
diff --git a/source/use_case/asr/src/UseCaseHandler.cc b/source/use_case/asr/src/UseCaseHandler.cc
index afcb6e4..7bce2c6 100644
--- a/source/use_case/asr/src/UseCaseHandler.cc
+++ b/source/use_case/asr/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -22,6 +22,7 @@
#include "hal.h"
#include "Wav2LetterMfcc.hpp"
#include "AudioUtils.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "AsrResult.hpp"
#include "Wav2LetterPreprocess.hpp"
diff --git a/source/use_case/img_class/src/UseCaseHandler.cc b/source/use_case/img_class/src/UseCaseHandler.cc
index fafc6b9..1f1d78b 100644
--- a/source/use_case/img_class/src/UseCaseHandler.cc
+++ b/source/use_case/img_class/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -19,6 +19,7 @@
#include "Classifier.hpp"
#include "InputFiles.hpp"
#include "MobileNetModel.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "hal.h"
#include "log_macros.h"
@@ -136,7 +137,7 @@
arm::app::DumpTensor(outputTensor);
#endif /* VERIFY_TEST_OUTPUT */
- if (!image::PresentInferenceResult(platform, results)) {
+ if (!PresentInferenceResult(platform, results)) {
return false;
}
diff --git a/source/use_case/kws/src/UseCaseHandler.cc b/source/use_case/kws/src/UseCaseHandler.cc
index c2d2ea4..8dd7724 100644
--- a/source/use_case/kws/src/UseCaseHandler.cc
+++ b/source/use_case/kws/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -22,6 +22,7 @@
#include "hal.h"
#include "MicroNetKwsMfcc.hpp"
#include "AudioUtils.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "KwsResult.hpp"
#include "log_macros.h"
diff --git a/source/use_case/kws_asr/src/UseCaseHandler.cc b/source/use_case/kws_asr/src/UseCaseHandler.cc
index bfc1d25..d598de6 100644
--- a/source/use_case/kws_asr/src/UseCaseHandler.cc
+++ b/source/use_case/kws_asr/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -19,6 +19,7 @@
#include "hal.h"
#include "InputFiles.hpp"
#include "AudioUtils.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "MicroNetKwsModel.hpp"
#include "MicroNetKwsMfcc.hpp"
diff --git a/source/use_case/noise_reduction/src/UseCaseHandler.cc b/source/use_case/noise_reduction/src/UseCaseHandler.cc
index 0c5984c..792b460 100644
--- a/source/use_case/noise_reduction/src/UseCaseHandler.cc
+++ b/source/use_case/noise_reduction/src/UseCaseHandler.cc
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2021 Arm Limited. All rights reserved.
+ * Copyright (c) 2021-2022 Arm Limited. All rights reserved.
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the "License");
@@ -18,6 +18,7 @@
#include "UseCaseHandler.hpp"
#include "UseCaseCommonUtils.hpp"
#include "AudioUtils.hpp"
+#include "ImageUtils.hpp"
#include "InputFiles.hpp"
#include "RNNoiseModel.hpp"
#include "RNNoiseProcess.hpp"
diff --git a/source/use_case/object_detection/include/DetectorPostProcessing.hpp b/source/use_case/object_detection/include/DetectorPostProcessing.hpp
index 3e9c819..5393f89 100644
--- a/source/use_case/object_detection/include/DetectorPostProcessing.hpp
+++ b/source/use_case/object_detection/include/DetectorPostProcessing.hpp
@@ -18,6 +18,7 @@
#define DETECTOR_POST_PROCESSING_HPP
#include "UseCaseCommonUtils.hpp"
+#include "ImageUtils.hpp"
#include "DetectionResult.hpp"
#include "YoloFastestModel.hpp"
@@ -45,20 +46,6 @@
int topN;
};
-
- struct Box {
- float x;
- float y;
- float w;
- float h;
- };
-
- struct Detection {
- Box bbox;
- std::vector<float> prob;
- float objectness;
- };
-
/**
* @brief Helper class to manage tensor post-processing for "object_detection"
* output.
@@ -99,18 +86,11 @@
int m_topN; /* TopN */
/**
- * @brief Calculate the Sigmoid function of the give value.
- * @param[in] x Value.
- * @return Sigmoid value of the input.
- **/
- float Sigmoid(float x);
-
- /**
* @brief Insert the given Detection in the list.
* @param[in] detections List of detections.
* @param[in] det Detection to be inserted.
**/
- void InsertTopNDetections(std::forward_list<Detection>& detections, Detection& det);
+ void InsertTopNDetections(std::forward_list<image::Detection>& detections, image::Detection& det);
/**
* @brief Given a Network calculate the detection boxes.
@@ -124,49 +104,7 @@
int imageWidth,
int imageHeight,
float threshold,
- std::forward_list<Detection>& detections);
-
- /**
- * @brief Calculate the 1D overlap.
- * @param[in] x1Center First center point.
- * @param[in] width1 First width.
- * @param[in] x2Center Second center point.
- * @param[in] width2 Second width.
- * @return The overlap between the two lines.
- **/
- float Calculate1DOverlap(float x1Center, float width1, float x2Center, float width2);
-
- /**
- * @brief Calculate the intersection between the two given boxes.
- * @param[in] box1 First box.
- * @param[in] box2 Second box.
- * @return The intersection value.
- **/
- float CalculateBoxIntersect(Box& box1, Box& box2);
-
- /**
- * @brief Calculate the union between the two given boxes.
- * @param[in] box1 First box.
- * @param[in] box2 Second box.
- * @return The two given boxes union value.
- **/
- float CalculateBoxUnion(Box& box1, Box& box2);
- /**
- * @brief Calculate the intersection over union between the two given boxes.
- * @param[in] box1 First box.
- * @param[in] box2 Second box.
- * @return The intersection over union value.
- **/
- float CalculateBoxIOU(Box& box1, Box& box2);
-
- /**
- * @brief Calculate the Non-Maxima suppression on the given detection boxes.
- * @param[in] detections Detection boxes.
- * @param[in] classes Number of classes.
- * @param[in] iouThreshold Intersection over union threshold.
- * @return true or false based on execution success.
- **/
- void CalculateNMS(std::forward_list<Detection>& detections, int classes, float iouThreshold);
+ std::forward_list<image::Detection>& detections);
/**
* @brief Draw on the given image a bounding box starting at (boxX, boxY).
diff --git a/source/use_case/object_detection/src/DetectorPostProcessing.cc b/source/use_case/object_detection/src/DetectorPostProcessing.cc
index edfb137..e97e6b3 100644
--- a/source/use_case/object_detection/src/DetectorPostProcessing.cc
+++ b/source/use_case/object_detection/src/DetectorPostProcessing.cc
@@ -15,6 +15,7 @@
* limitations under the License.
*/
#include "DetectorPostProcessing.hpp"
+#include "PlatformMath.hpp"
#include <algorithm>
#include <cmath>
@@ -75,7 +76,7 @@
int originalImageWidth = originalImageSize;
int originalImageHeight = originalImageSize;
- std::forward_list<Detection> detections;
+ std::forward_list<image::Detection> detections;
GetNetworkBoxes(net, originalImageWidth, originalImageHeight, m_threshold, detections);
/* Do nms */
@@ -124,15 +125,11 @@
}
}
-float DetectorPostprocessing::Sigmoid(float x)
-{
- return 1.f/(1.f + exp(-x));
-}
-void DetectorPostprocessing::InsertTopNDetections(std::forward_list<Detection>& detections, Detection& det)
+void DetectorPostprocessing::InsertTopNDetections(std::forward_list<image::Detection>& detections, image::Detection& det)
{
- std::forward_list<Detection>::iterator it;
- std::forward_list<Detection>::iterator last_it;
+ std::forward_list<image::Detection>::iterator it;
+ std::forward_list<image::Detection>::iterator last_it;
for ( it = detections.begin(); it != detections.end(); ++it ) {
if(it->objectness > det.objectness)
break;
@@ -144,11 +141,11 @@
}
}
-void DetectorPostprocessing::GetNetworkBoxes(Network& net, int imageWidth, int imageHeight, float threshold, std::forward_list<Detection>& detections)
+void DetectorPostprocessing::GetNetworkBoxes(Network& net, int imageWidth, int imageHeight, float threshold, std::forward_list<image::Detection>& detections)
{
int numClasses = net.numClasses;
int num = 0;
- auto det_objectness_comparator = [](Detection& pa, Detection& pb) {
+ auto det_objectness_comparator = [](image::Detection& pa, image::Detection& pb) {
return pa.objectness < pb.objectness;
};
for (size_t i = 0; i < net.branches.size(); ++i) {
@@ -162,10 +159,10 @@
/* Objectness score */
int bbox_obj_offset = h * width * channel + w * channel + anc * (numClasses + 5) + 4;
- float objectness = Sigmoid(((float)net.branches[i].modelOutput[bbox_obj_offset] - net.branches[i].zeroPoint) * net.branches[i].scale);
+ float objectness = math::MathUtils::SigmoidF32(((float)net.branches[i].modelOutput[bbox_obj_offset] - net.branches[i].zeroPoint) * net.branches[i].scale);
if(objectness > threshold) {
- Detection det;
+ image::Detection det;
det.objectness = objectness;
/* Get bbox prediction data for each anchor, each feature point */
int bbox_x_offset = bbox_obj_offset -4;
@@ -183,8 +180,8 @@
float bbox_x, bbox_y;
/* Eliminate grid sensitivity trick involved in YOLOv4 */
- bbox_x = Sigmoid(det.bbox.x);
- bbox_y = Sigmoid(det.bbox.y);
+ bbox_x = math::MathUtils::SigmoidF32(det.bbox.x);
+ bbox_y = math::MathUtils::SigmoidF32(det.bbox.y);
det.bbox.x = (bbox_x + w) / width;
det.bbox.y = (bbox_y + h) / height;
@@ -192,7 +189,7 @@
det.bbox.h = exp(det.bbox.h) * net.branches[i].anchor[anc*2+1] / net.inputHeight;
for (int s = 0; s < numClasses; s++) {
- float sig = Sigmoid(((float)net.branches[i].modelOutput[bbox_scores_offset + s] - net.branches[i].zeroPoint) * net.branches[i].scale)*objectness;
+ float sig = math::MathUtils::SigmoidF32(((float)net.branches[i].modelOutput[bbox_scores_offset + s] - net.branches[i].zeroPoint) * net.branches[i].scale)*objectness;
det.prob.emplace_back((sig > threshold) ? sig : 0);
}
@@ -221,81 +218,6 @@
num -=1;
}
-float DetectorPostprocessing::Calculate1DOverlap(float x1Center, float width1, float x2Center, float width2)
-{
- float left_1 = x1Center - width1/2;
- float left_2 = x2Center - width2/2;
- float leftest = left_1 > left_2 ? left_1 : left_2;
-
- float right_1 = x1Center + width1/2;
- float right_2 = x2Center + width2/2;
- float rightest = right_1 < right_2 ? right_1 : right_2;
-
- return rightest - leftest;
-}
-
-float DetectorPostprocessing::CalculateBoxIntersect(Box& box1, Box& box2)
-{
- float width = Calculate1DOverlap(box1.x, box1.w, box2.x, box2.w);
- if (width < 0) {
- return 0;
- }
- float height = Calculate1DOverlap(box1.y, box1.h, box2.y, box2.h);
- if (height < 0) {
- return 0;
- }
-
- float total_area = width*height;
- return total_area;
-}
-
-float DetectorPostprocessing::CalculateBoxUnion(Box& box1, Box& box2)
-{
- float boxes_intersection = CalculateBoxIntersect(box1, box2);
- float boxes_union = box1.w * box1.h + box2.w * box2.h - boxes_intersection;
- return boxes_union;
-}
-
-
-float DetectorPostprocessing::CalculateBoxIOU(Box& box1, Box& box2)
-{
- float boxes_intersection = CalculateBoxIntersect(box1, box2);
- if (boxes_intersection == 0) {
- return 0;
- }
-
- float boxes_union = CalculateBoxUnion(box1, box2);
- if (boxes_union == 0) {
- return 0;
- }
-
- return boxes_intersection / boxes_union;
-}
-
-void DetectorPostprocessing::CalculateNMS(std::forward_list<Detection>& detections, int classes, float iouThreshold)
-{
- int idxClass{0};
- auto CompareProbs = [idxClass](Detection& prob1, Detection& prob2) {
- return prob1.prob[idxClass] > prob2.prob[idxClass];
- };
-
- for (idxClass = 0; idxClass < classes; ++idxClass) {
- detections.sort(CompareProbs);
-
- for (std::forward_list<Detection>::iterator it=detections.begin(); it != detections.end(); ++it) {
- if (it->prob[idxClass] == 0) continue;
- for (std::forward_list<Detection>::iterator itc=std::next(it, 1); itc != detections.end(); ++itc) {
- if (itc->prob[idxClass] == 0) {
- continue;
- }
- if (CalculateBoxIOU(it->bbox, itc->bbox) > iouThreshold) {
- itc->prob[idxClass] = 0;
- }
- }
- }
- }
-}
-
void DetectorPostprocessing::DrawBoxOnImage(uint8_t* imgIn, int imWidth, int imHeight, int boxX,int boxY, int boxWidth, int boxHeight)
{
auto CheckAndFixOffset = [](int im_width,int im_height,int& offset) {
diff --git a/source/use_case/vww/src/UseCaseHandler.cc b/source/use_case/vww/src/UseCaseHandler.cc
index 01011e2..a47f191 100644
--- a/source/use_case/vww/src/UseCaseHandler.cc
+++ b/source/use_case/vww/src/UseCaseHandler.cc
@@ -18,6 +18,7 @@
#include "VisualWakeWordModel.hpp"
#include "Classifier.hpp"
#include "InputFiles.hpp"
+#include "ImageUtils.hpp"
#include "UseCaseCommonUtils.hpp"
#include "hal.h"
#include "log_macros.h"
@@ -144,7 +145,7 @@
arm::app::DumpTensor(outputTensor);
#endif /* VERIFY_TEST_OUTPUT */
- if (!image::PresentInferenceResult(platform, results)) {
+ if (!PresentInferenceResult(platform, results)) {
return false;
}