a1997-dom / object-detection-with-d435-and-yolov3

A target recognition and location algorithm using realsense D435 and yolov3

Object-Detection-and-location-RealsenseD435 and yolov3

Requirements

Ubuntu18.04 OR 16.04
Opencv 4.x
C++ 11_std At least,I used the C++ 17 std
Eigen3 :in absolutely Path /usr/local/eigen3
Cmake>= 3.17
PCL lib>=1.7.1
Intel Realsense SDK >=2.0
Yolov3

How to use

git clone https://github.com/a1997-dom/Object-Detection-with-D435-and-yolov3.git
cd  Object-Detection-with-D435-and-yolov3/DNN/engine/
wget https://pjreddie.com/media/files/yolov3.weights ;wget https://pjreddie.com/media/files/yolov3-tiny.weights

you Can change the engine path in src/main.cpp
on line 25-27

String yolo_tiny_model ="../engine/yolov3.weights";
String yolo_tiny_cfg =  "../engine/yolov3.cfg";
String classname_path="../engine/coco.names";

You can use your weight by Darknet or others supported by DNN too

cd ..
mkdir bulid; cd build
cmake ..
make
./DNN_Yolo

Attention:Default parameter on line 251 and 252 in src/main.cpp

    net.setPreferableBackend(DNN_BACKEND_OPENCV);// DNN_BACKEND_INFERENCE_ENGINE DNN_BACKEND_CUDA
    net.setPreferableTarget(DNN_TARGET_CPU);//DNN_TARGET_CUDA

if you have IntelCore CPU you can chose "DNN_BACKEND_INFERENCE_ENGINE"to accelerate youe model--Openvino;
But you should make sure your CPU is Intel and the Contrib of Opencv has been installed.
If you have GPU(From Nvidia),You can Think about The Cuda acceleration.Before this you should reinstall Your Opencv(Version Most>4.2) with This:OpenCV_DNN
Open the Cuda Setting when CMake.