video-tracker

Description

Application for tracking various objects and estimating their speed in video stream.

Installation

Clone repository and all submodules

• git clone https://github.com/vnkrtv/video-tracker.git && cd video-tracker
• git pull --recurse-submodules
• git submodule init
• git submodule update --remote --recursive

OpenCV installation

Installing all system requirements described in this article
After installing them is's able to build OpenCV

• cd contrib/opencv
• mkdir build && cd build
• cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D WITH_GSTREAMER=ON -D WITH_FFMPEG=ON [-D WITH_CUDA=ON -D ENABLE_FAST_MATH=1 -D CUDA_FAST_MATH=1 -D WITH_CUBLAS=1]..
• make [-j 9]
• sudo make install

dlib installation

• cd contrib/dlib/dlib
• mkdir build && cd build
• cmake -D CMAKE_BUILD_TYPE=RELEASE -D BUILD_SHARED_LIBS=1 -D USE_AVX_INSTRUCTIONS=ON [-D DLIB_USE_CUDA=1] ..
• make [-j 9]
• sudo make install

Building application

• cmake -DCMAKE_BUILD_TYPE=RELEASE .
• cmake --build cmake-build-release --target video_tracker [-- -j 9]

Usage

 Options: 

                 -h, --help Show help  
   --video-src, -v [string] Video source (video file, ip camera, video device)  
  --model-path, -m [string] MobileNetSSD folder path  
      --output, -o [string] Output file name. By default, processed video stream is not 
                            saving  
 --classes, -c [integer...] Set of detected classes ID. Full set could be found 
                            in README. Default classes: persons and cars  
  --confidence, -t [number] Model's confidence coefficient. Default value: 0.4  
                --no-window Does not show named window with video stream. False 
                            by default  
                     --cuda Use GPU with CUDA  

Model

MobileNet is using in project for objects detection. Model is pre-trained and taken from https://github.com/chuanqi305/MobileNet-SSD//. It was trained in Caffe-SSD framework. This model can detect 20 classes. Available classes can be found in table below:

Class name	Class ID
background	0
aeroplane	1
bicycle	2
bird	3
boat	4
bottle	5
bus	6
car	7
cat	8
chair	9
cow	10
dining table	11
dog	12
horse	13
motorbike	14
person	15
potted plant	16
sheep	17
sofa	18
train	19
tv monitor	20

To make application detect multiple classes, you need specify special -c, --classes flag. Example:

• video_tracker --video-src /dev/video0 --model-path MobileNetSSD --classes {8,12} - in this case app will detectObjects only cats and dogs using camera /dev/video0