Skip to content
The open-sourced software for place recognition using multiple descriptors derived from multi-modal images.
C++ C
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Open Multimodal Place Recognition

How to run the code?

Prepare the environment and the dependencies on your computer, then download the source code, the additional files and the dataset. Before running the code, you may change the settings in the configuration file.

The GUI of OpenMPR.

Related Paper

If you are using this code in your research, please cite the paper:
Cheng, Ruiqi, et al. "OpenMPR: Recognize places using multimodal data for people with visual impairments." Measurement Science and Technology (2019).


Visual Studio 2017 on Windows 10


OpenCV 4.0 (64 bit): set OpenCV path as OpenCV_DIR in system environments
DBoW3: set DBoW3 path as DBoW3_DIR in system environments
FFTW3: set FFTW path as FFTW_DIR in system environments

Additional Files

The pre-trained model of CNN could be downloaded at GoogLeNet-Places365, which should be unzipped to the source code folder.


The dataset is available at Multimodal Dataset.

Configuration File

The configuration file Config.yaml is in the folder of OpenMultiPR. The detailed information of the parameters could be found in Config.yaml.

In the yaml file, the dataset and BoW vocabulary paths are assigned. The configuration files also includes the parameters on wheter to use the specific modalities (i.e. RGB, Infrared, Depth and GNSS data) and the corresponding descriptors (i.e. GIST, ORB-BoW, LDB). The running mode could be set in the file.

You can’t perform that action at this time.