an implementation of Yolov4 algorithm to ROS using darknet framework and a python wrapper (PyYOLO)
In detection.py configure your cfg, data and weight file locations on those lines;
detector = pyyolo.YOLO(pkg_path+"/src/cfg/yolov4.cfg", # cfg file location
pkg_path+"/src/yolov4.weights", # weight file location
pkg_path+"/src/cfg/coco.data", # data file location
detection_threshold = 0.5,
hier_threshold = 0.5,
nms_threshold = 0.45)
Then add detection.py as node to your launch file as seen in example here. Don't forget to remap "camera/data" to source Image topic and detection/output to desired output Image topic. All detections will be also published on object_detection/detections in a custom message format I named as object_detection. You may need to create and generate that message type before you publish that. Or just comment out those in detection.py;
#line 25 - 29
detection.min_x, detection.min_y, detection.max_x, detection.max_y = det.to_xyxy()
detection.probability = det.prob
detection.class_ = det.name
detection.source = topic
detection_pub.publish(detection)
#line 50
global detection_pub
#line 66
detection_pub = rospy.Publisher("object_detection/detection",object_detection,queue_size=0)
object_detection.msg
int16 min_x
int16 max_x
int16 min_y
int16 max_y
string class
float32 probability
string source
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent