yolo3 flir ok

CMakeLists.txt

@@ -89,6 +89,9 @@ target_link_libraries(test_yolo3_coco4 tkDNN)
 
 add_executable(test_yolo3_berkeley tests/yolo3_berkeley/yolo3_berkeley.cpp)
 target_link_libraries(test_yolo3_berkeley tkDNN)
+
+add_executable(test_yolo3_flir tests/yolo3_flir/yolo3_flir.cpp)
+target_link_libraries(test_yolo3_flir tkDNN)
 ################################################################################
 
 

demo/demo/demo.cpp

@@ -12,7 +12,9 @@
 #include "Yolo3Detection.h"
 
 bool 			gRun;
-std::string   	obj_class[10] {"person", "car", "truck", "bus", "motor", "bike", "rider", "traffic light", "traffic sign", "train"};
+//std::string   	obj_class[10] {"person", "car", "truck", "bus", "motor", "bike", "rider", "traffic light", "traffic sign", "train"};
+//std::string   	obj_class[3] {"person", "bike", "car"};
+std::string   	obj_class[10] {"0", "1", "2", "3", "4", "5", "6", "7", "8", "9"};
 
 
 void sig_handler(int signo) {

src/Yolo3Detection.cpp

@@ -68,18 +68,19 @@ void Yolo3Detection::update(cv::Mat &imageORIG) {
     float yRatio =  float(imageORIG.rows) / float(netRT->input_dim.h);
 
     resize(imageORIG, imageORIG, cv::Size(netRT->input_dim.w, netRT->input_dim.h));
+
     imageORIG.convertTo(imageF, CV_32FC3, 1/255.0); 
 
     //split channels
     cv::split(imageF,bgr);//split source
 
     //write channels
-    int idx = 0;
-    memcpy((void*)&input[idx], (void*)bgr[2].data, imageF.rows*imageF.cols*sizeof(dnnType));
-    idx = imageF.rows*imageF.cols;
-    memcpy((void*)&input[idx], (void*)bgr[1].data, imageF.rows*imageF.cols*sizeof(dnnType));
-    idx *= 2;    
-    memcpy((void*)&input[idx], (void*)bgr[0].data, imageF.rows*imageF.cols*sizeof(dnnType));
+    for(int i=0; i<netRT->input_dim.c; i++) {
+        int idx = i*imageF.rows*imageF.cols;
+        int ch = netRT->input_dim.c-1 -i;
+        memcpy((void*)&input[idx], (void*)bgr[ch].data, imageF.rows*imageF.cols*sizeof(dnnType));
+    }
+
 
     //DO INFERENCE
     dnnType *rt_out[3]; 

tests/yolo3_flir/yolo3_flir.cfg

@@ -17,7 +17,7 @@ hue=.1
 
 learning_rate=0.001
 burn_in=1000
-max_batches = 10000
+max_batches = 20000
 policy=steps
 steps=8000,9000
 scales=.1,.1
@@ -602,13 +602,13 @@ activation=leaky
 size=1
 stride=1
 pad=1
-filters=30
+filters=24
 activation=linear
 
 [yolo]
 mask = 6,7,8
 anchors = 8.2087,8.5515, 18.4134,20.3391, 40.2194,29.2990, 31.6137,69.2240, 69.8497,48.3838, 108.8817,76.6316, 96.5753,145.5743, 165.9182,117.4493, 215.7497,198.4648
-classes=5
+classes=3
 num=9
 jitter=.3
 ignore_thresh = .5
@@ -686,13 +686,13 @@ activation=leaky
 size=1
 stride=1
 pad=1
-filters=30
+filters=24
 activation=linear
 
 [yolo]
 mask = 3,4,5
 anchors = 8.2087,8.5515, 18.4134,20.3391, 40.2194,29.2990, 31.6137,69.2240, 69.8497,48.3838, 108.8817,76.6316, 96.5753,145.5743, 165.9182,117.4493, 215.7497,198.4648
-classes=5
+classes=3
 num=9
 jitter=.3
 ignore_thresh = .5
@@ -770,13 +770,13 @@ activation=leaky
 size=1
 stride=1
 pad=1
-filters=30
+filters=24
 activation=linear
 
 [yolo]
 mask = 0,1,2
 anchors = 8.2087,8.5515, 18.4134,20.3391, 40.2194,29.2990, 31.6137,69.2240, 69.8497,48.3838, 108.8817,76.6316, 96.5753,145.5743, 165.9182,117.4493, 215.7497,198.4648
-classes=5
+classes=3
 num=9
 jitter=.3
 ignore_thresh = .5

tests/yolo3_flir/yolo3_flir.cpp

@@ -11,7 +11,7 @@ int main() {
 
     // create yolo3 model
     std::string bin_path  = "../tests/yolo3_flir";
-    int classes = 5;
+    int classes = 3;
     tk::dnn::Yolo *yolo [3];
     #include "models/Yolo3.h"