DeployedDetector

Introduce

The Project includes three parts in total.

1. one is for postprocess for detector of ssd and centernet
2. the other is for instance manager, managing all objects detected from start time and calculate depth, location, get all visible objects, pass them on to semantic map
3. the last one is for samrpn tracking algorithm.

Installation

requirements

1. opencv, build it from source

2. Eigen3

   sudo apt install libeigen3-dev

3. glog, build it from source

4. MNN inference framework git clone mnn lib from github, compile it and generate lib and include headers put them in MNN/ directory

compile

# build detector
mkdir build && cd build && cmake .. && make -j`nproc`

# if want to use sdk, use instead
cmake .. -DBUILD_SDK=ON

# build tracker
cmake .. -DBUILD_TRACKER=ON

Development

please refer to src/detect.cpp for example, if you want to change detector type(ssd or centernet)

// centernet detector
detector.reset(new CenterNetDetector(model_name));

// ssd detector
detector.reset(new Detector(model_name));