From 331d327760e641f32f9813ddc9708addaf5d065e Mon Sep 17 00:00:00 2001
From: Gursimar Singh <gursimar@bigvisionllc.com>
Date: Mon, 28 Oct 2024 12:19:33 +0530
Subject: [PATCH] Merge pull request #26336 from
 gursimarsingh:person_reid_bug_fix

[BUG FIX] Fix issues in Person ReID C++ sample #26336

This PR fixes multiple issues in the Person ReID C++ sample that were causing incorrect outputs. It addresses improper matrix initialization, adds a missing return statement, and ensures that vectors are properly cleared before reuse. These changes correct the output of the sample.
### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [x] The PR is proposed to the proper branch
- [x] There is a reference to the original bug report and related work
- [x] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [x] The feature is well documented and sample code can be built with the project CMake
---
 samples/dnn/person_reid.cpp | 16 ++++++++++++----
 samples/dnn/person_reid.py  |  7 +++++--
 2 files changed, 17 insertions(+), 6 deletions(-)

diff --git a/samples/dnn/person_reid.cpp b/samples/dnn/person_reid.cpp
index 506e7446f5..3f128df9f0 100644
--- a/samples/dnn/person_reid.cpp
+++ b/samples/dnn/person_reid.cpp
@@ -107,10 +107,11 @@ static void extractFeatures(vector<Mat> &imglist, Net &net, vector<Mat> &feature
         Mat out=net.forward();
         vector<int> s {out.size[0], out.size[1]};
         out = out.reshape(1, s);
-        features.resize(out.rows);
         for (int i = 0; i < out.rows; i++)
         {
-            normalize(out.row(i), features[i], 1.0, 0.0, NORM_L2);
+            Mat norm_features;
+            normalize(out.row(i), norm_features, 1.0, 0.0, NORM_L2);
+            features.push_back(norm_features);
         }
     }
     return;
@@ -212,6 +213,7 @@ static void yoloDetector(Mat &frame, Net &net, vector<Mat>& images)
         images[i] = frame(roi); // Crop the region from the frame
         imgDict[images[i]] = roi;
     }
+    return;
 }
 
 int main(int argc, char **argv)
@@ -256,7 +258,11 @@ int main(int argc, char **argv)
     int fontSize = 50;
     int fontWeight = 500;
 
-    Net reidNet = readNetFromONNX(modelPath);
+    EngineType engine = ENGINE_AUTO;
+    if (backend != "default" || target != "cpu"){
+        engine = ENGINE_CLASSIC;
+    }
+    Net reidNet = readNetFromONNX(modelPath, engine);
     reidNet.setPreferableBackend(getBackendID(backend));
     reidNet.setPreferableTarget(getTargetID(target));
 
@@ -264,7 +270,7 @@ int main(int argc, char **argv)
         cout<<"[ERROR] Please pass path to yolov8.onnx model file using --yolo_model."<<endl;
         return -1;
     }
-    Net net = readNetFromONNX(yoloPath);
+    Net net = readNetFromONNX(yoloPath, engine);
 
     FontFace fontFace("sans");
 
@@ -342,6 +348,8 @@ int main(int argc, char **argv)
             fontSize = min(fontSize, (stdSize*imgWidth)/stdImgSize);
             fontWeight = min(fontWeight, (stdWeight*imgWidth)/stdImgSize);
         }
+        detectedImages.clear();
+        galleryFeatures.clear();
 
         yoloDetector(frame, net, detectedImages);
         extractFeatures(detectedImages, reidNet, galleryFeatures);
diff --git a/samples/dnn/person_reid.py b/samples/dnn/person_reid.py
index 48fd0d28e6..4f63268e4f 100644
--- a/samples/dnn/person_reid.py
+++ b/samples/dnn/person_reid.py
@@ -183,8 +183,11 @@ def main():
     else:
         args.yolo_model = findModel(args.yolo_model, args.yolo_sha1)
 
-    yolo_net = cv.dnn.readNet(args.yolo_model)
-    reid_net = cv.dnn.readNet(args.model)
+    engine = cv.dnn.ENGINE_AUTO
+    if args.backend != "default" or args.backend != "cpu":
+        engine = cv.dnn.ENGINE_CLASSIC
+    yolo_net = cv.dnn.readNetFromONNX(args.yolo_model, engine)
+    reid_net = cv.dnn.readNetFromONNX(args.model, engine)
     reid_net.setPreferableBackend(get_backend_id(args.backend))
     reid_net.setPreferableTarget(get_target_id(args.target))
     cap = cv.VideoCapture(cv.samples.findFile(args.input) if args.input else 0)