Skip to content

Deepstream application that performs face detection using retinanet and multi object tracking using deepsort

Notifications You must be signed in to change notification settings

bqm1111/deepstream-retinanet

Repository files navigation

Dependency library

Install dependency library: rapidjson, json-glib-1.0, eigen3

sudo apt-get update
sudo apt-get install -y rapidjson-dev libjson-glib-dev libeigen3-dev

Convert model to .trt file

cd tools
sh convert.sh

Run

cd build
cmake ..
make -j
./experiment <list_video_source_file> <is_live>

Run with video source file

./experiment ../configs/video_list.json 0

Run with rtsp stream

./experiment ../configs/rtsp_source_list.json 1

Enable LATENCY mesurement

export NVDS_ENABLE_COMPONENT_LATENCY_MEASUREMENT=1

Check memory with LeakSanitizer

Add the following flag option to enable LeakSanitizer tool

set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -fno-omit-frame-pointer -fsanitize=leak -g -pthread")

Before running the program to check memory leak, remember to add the following enviroment variable

export LD_PRELOAD=<path-to-libgstnvfacealign.so>

Run with docker

docker build --build-arg USER_UID=1003,USER_GID=999 . --tag deepstream-app
docker run -ti --rm -v deepstream-engine:/workspace/build --name deepstream-app deepstream-app
docker exec -it deepstream-app bash

About

Deepstream application that performs face detection using retinanet and multi object tracking using deepsort

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published