kunle12 / dn_object_detect

Multiclass object detection ROS node based on YOLO/Darknet

Multi-class object detection with darknet/yolo

Introduction

This is a ROS node for muli-class object detection using darknet/yolo Deep Neural Network (DNN).

Compilation

Fully check out this repository with submodule darknet. Compile the darknet library with the following commands:

cd darknet;mkdir build;cd build
cmake -DCMAKE_INSTALL_PREFIX=<full path to dn_object_detect> ..
make install

NOTE If you are compiling darknet submodule with OpenCV 3.x, you need to add the following lines in opencv2/core/types_c.h after #include "opencv2/core/cvdef.h":

#ifndef __cplusplus
#include "opencv2/core/fast_math.hpp"
#endif

Go back to the catkin workspace base directory and compile the ROS node with

catkin_make dn_object_detect

Running Node

NOTE You need to download the model/weights file separately.

roslaunch dn_object_detect objdetect.launch

The node will publish the following two topics

/dn_object_detect/detected_objects detected object list.

/dn_object_detect/debug_view debugging image stream.