Skip to content

Commit a42ca2e

Browse files
alalekalalek
committed
Merge pull request opencv#10483 from dkurt:fix_dnn_ssd_cpp_sample
2 parents bf80c94 + d0580df commit a42ca2e

File tree

1 file changed

+31
-57
lines changed

1 file changed

+31
-57
lines changed

‎samples/dnn/ssd_mobilenet_object_detection.cpp

Lines changed: 31 additions & 57 deletions
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,6 @@ using namespace std;
1313

1414
const size_t inWidth = 300;
1515
const size_t inHeight = 300;
16-
const float WHRatio = inWidth / (float)inHeight;
1716
const float inScaleFactor = 0.007843f;
1817
const float meanVal = 127.5;
1918
const char* classNames[] = {"background",
@@ -23,13 +22,6 @@ const char* classNames[] = {"background",
2322
"motorbike", "person", "pottedplant",
2423
"sheep", "sofa", "train", "tvmonitor"};
2524

26-
const char* about = "This sample uses MobileNet Single-Shot Detector "
27-
"(https://arxiv.org/abs/1704.04861) "
28-
"to detect objects on camera/video/image.\n"
29-
".caffemodel model's file is available here: "
30-
"https://github.com/chuanqi305/MobileNet-SSD\n"
31-
"Default network is 300x300 and 20-classes VOC.\n";
32-
3325
const char* params
3426
= "{ help | false | print usage }"
3527
"{ proto | MobileNetSSD_deploy.prototxt | model configuration }"
@@ -44,16 +36,22 @@ const char* params
4436
int main(int argc, char** argv)
4537
{
4638
CommandLineParser parser(argc, argv, params);
47-
48-
if (parser.get<bool>("help"))
39+
parser.about("This sample uses MobileNet Single-Shot Detector "
40+
"(https://arxiv.org/abs/1704.04861) "
41+
"to detect objects on camera/video/image.\n"
42+
".caffemodel model's file is available here: "
43+
"https://github.com/chuanqi305/MobileNet-SSD\n"
44+
"Default network is 300x300 and 20-classes VOC.\n");
45+
46+
if (parser.get<bool>("help") || argc == 1)
4947
{
50-
cout << about << endl;
5148
parser.printMessage();
5249
return 0;
5350
}
5451

5552
String modelConfiguration = parser.get<string>("proto");
5653
String modelBinary = parser.get<string>("model");
54+
CV_Assert(!modelConfiguration.empty() && !modelBinary.empty());
5755

5856
//! [Initialize network]
5957
dnn::Net net = readNetFromCaffe(modelConfiguration, modelBinary);
@@ -75,7 +73,7 @@ int main(int argc, char** argv)
7573
}
7674

7775
VideoCapture cap;
78-
if (parser.get<String>("video").empty())
76+
if (!parser.has("video"))
7977
{
8078
int cameraDevice = parser.get<int>("camera_device");
8179
cap = VideoCapture(cameraDevice);
@@ -95,32 +93,16 @@ int main(int argc, char** argv)
9593
}
9694
}
9795

98-
Size inVideoSize;
99-
inVideoSize = Size((int) cap.get(CV_CAP_PROP_FRAME_WIDTH), //Acquire input size
100-
(int) cap.get(CV_CAP_PROP_FRAME_HEIGHT));
101-
102-
Size cropSize;
103-
if (inVideoSize.width / (float)inVideoSize.height > WHRatio)
104-
{
105-
cropSize = Size(static_cast<int>(inVideoSize.height * WHRatio),
106-
inVideoSize.height);
107-
}
108-
else
109-
{
110-
cropSize = Size(inVideoSize.width,
111-
static_cast<int>(inVideoSize.width / WHRatio));
112-
}
113-
114-
Rect crop(Point((inVideoSize.width - cropSize.width) / 2,
115-
(inVideoSize.height - cropSize.height) / 2),
116-
cropSize);
96+
//Acquire input size
97+
Size inVideoSize((int) cap.get(CV_CAP_PROP_FRAME_WIDTH),
98+
(int) cap.get(CV_CAP_PROP_FRAME_HEIGHT));
11799

118100
double fps = cap.get(CV_CAP_PROP_FPS);
119101
int fourcc = static_cast<int>(cap.get(CV_CAP_PROP_FOURCC));
120102
VideoWriter outputVideo;
121103
outputVideo.open(parser.get<String>("out") ,
122104
(fourcc != 0 ? fourcc : VideoWriter::fourcc('M','J','P','G')),
123-
(fps != 0 ? fps : 10.0), cropSize, true);
105+
(fps != 0 ? fps : 10.0), inVideoSize, true);
124106

125107
for(;;)
126108
{
@@ -138,15 +120,17 @@ int main(int argc, char** argv)
138120

139121
//! [Prepare blob]
140122
Mat inputBlob = blobFromImage(frame, inScaleFactor,
141-
Size(inWidth, inHeight), meanVal, false); //Convert Mat to batch of images
123+
Size(inWidth, inHeight),
124+
Scalar(meanVal, meanVal, meanVal),
125+
false, false); //Convert Mat to batch of images
142126
//! [Prepare blob]
143127

144128
//! [Set input blob]
145-
net.setInput(inputBlob, "data"); //set the network input
129+
net.setInput(inputBlob); //set the network input
146130
//! [Set input blob]
147131

148132
//! [Make forward pass]
149-
Mat detection = net.forward("detection_out"); //compute output
133+
Mat detection = net.forward(); //compute output
150134
//! [Make forward pass]
151135

152136
vector<double> layersTimings;
@@ -155,13 +139,10 @@ int main(int argc, char** argv)
155139

156140
Mat detectionMat(detection.size[2], detection.size[3], CV_32F, detection.ptr<float>());
157141

158-
frame = frame(crop);
159-
160-
ostringstream ss;
161142
if (!outputVideo.isOpened())
162143
{
163-
ss << "FPS: " << 1000/time << " ; time: " << time << " ms";
164-
putText(frame, ss.str(), Point(20,20), 0, 0.5, Scalar(0,0,255));
144+
putText(frame, format("FPS: %.2f ; time: %.2f ms", 1000.f/time, time),
145+
Point(20,20), 0, 0.5, Scalar(0,0,255));
165146
}
166147
else
167148
cout << "Inference time, ms: " << time << endl;
@@ -175,27 +156,20 @@ int main(int argc, char** argv)
175156
{
176157
size_t objectClass = (size_t)(detectionMat.at<float>(i, 1));
177158

178-
int xLeftBottom = static_cast<int>(detectionMat.at<float>(i, 3) * frame.cols);
179-
int yLeftBottom = static_cast<int>(detectionMat.at<float>(i, 4) * frame.rows);
180-
int xRightTop = static_cast<int>(detectionMat.at<float>(i, 5) * frame.cols);
181-
int yRightTop = static_cast<int>(detectionMat.at<float>(i, 6) * frame.rows);
182-
183-
ss.str("");
184-
ss << confidence;
185-
String conf(ss.str());
186-
187-
Rect object((int)xLeftBottom, (int)yLeftBottom,
188-
(int)(xRightTop - xLeftBottom),
189-
(int)(yRightTop - yLeftBottom));
159+
int left = static_cast<int>(detectionMat.at<float>(i, 3) * frame.cols);
160+
int top = static_cast<int>(detectionMat.at<float>(i, 4) * frame.rows);
161+
int right = static_cast<int>(detectionMat.at<float>(i, 5) * frame.cols);
162+
int bottom = static_cast<int>(detectionMat.at<float>(i, 6) * frame.rows);
190163

191-
rectangle(frame, object, Scalar(0, 255, 0));
192-
String label = String(classNames[objectClass]) + ": " + conf;
164+
rectangle(frame, Point(left, top), Point(right, bottom), Scalar(0, 255, 0));
165+
String label = format("%s: %.2f", classNames[objectClass], confidence);
193166
int baseLine = 0;
194167
Size labelSize = getTextSize(label, FONT_HERSHEY_SIMPLEX, 0.5, 1, &baseLine);
195-
rectangle(frame, Rect(Point(xLeftBottom, yLeftBottom - labelSize.height),
196-
Size(labelSize.width, labelSize.height + baseLine)),
168+
top = max(top, labelSize.height);
169+
rectangle(frame, Point(left, top - labelSize.height),
170+
Point(left + labelSize.width, top + baseLine),
197171
Scalar(255, 255, 255), CV_FILLED);
198-
putText(frame, label, Point(xLeftBottom, yLeftBottom),
172+
putText(frame, label, Point(left, top),
199173
FONT_HERSHEY_SIMPLEX, 0.5, Scalar(0,0,0));
200174
}
201175
}

0 commit comments

Comments
 (0)