@openhps/openvslam adds bindings and OpenHPS abstracts for [OpenVSLAM](https://stella-vslam.readthedocs.io/en/latest/). It supports [@openhps/opencv](https://github.com/OpenHPS/openhps-opencv) for video processing before sending it to OpenVSLAM. This module will create C++ bindings to OpenVSLAM and supports monocular and stereo SLAM (originally based on ORB-SLAM2).
- Install OpenCV 4.X-4.3.X (4.4.0 or higher is not supported)
- Install OpenVSLAM according to the instructions on https://openvslam-community.readthedocs.io/en/latest/ We have a docker container for the unit tests that you can investigate.
- You will need to declare the following environment variables before installing this NPM module
OPENCV_LIB_DIR: Library directory of OpenCVOPENCV_INCLUDE_DIR: Include directory of OpenCVOPENVSLAM_LIB_DIR: OpenVSLAM library directoryOPENVSLAM_INCLUDE_DIR: OpenVSLAM include directory
- Install @openhps/core if you have not installed it already
- Install @openhps/opencv if you have not installed it already
-
npm install @openhps/openvslam --save
OpenVSLAM uses a map database, vocabulary and configuration YAML file to start the mapping and/or localization. Basic usage allows you to specify the locations of these files to start the processing.
import { ModelBuilder, CallbackSinkNode } from '@openhps/core';
import { VideoSource, CameraObject } from '@openhps/opencv';
import { VSLAMProcessingNode } from '@openhps/openvslam';
ModelBuilder.create()
.from(new VideoSource({
source: new CameraObject(),
videoSource: "someVideo.mp4",
autoPlay: true
}))
.via(new VSLAMProcessingNode({
config: "/path/to/config.yaml", // OpenVSLAM camera configuration
mapping: true, // Enable to disable the mapping
mapDatabaseFile: "/path/to/map.msg", // Map database file in MessagePack format
vocabularyFile: "/path/to/vocab.fbow", // OpenVSLAM vocabulary file
persistMapping: true, // Persist any changes to the map data file
}))
.to(new CallbackSinkNode(frame => {
// frame.source.position
}))
.build();OpenVSLAM comes with a socket viewer web application that will visualize the landmarks, keyframes, FPS while also allowing you to control the processing from within your browser. OpenHPS provides the socket viewer as a sink node.
import { ModelBuilder } from '@openhps/core';
import { VideoSource } from '@openhps/opencv';
import { VSLAMProcessingNode, VSLAMSocketViewer } from '@openhps/openvslam';
ModelBuilder.create()
.from(new VideoSource({
// ...
}))
.via(new VSLAMProcessingNode({
// ...
}))
.to(new VSLAMSocketViewer({
url: "http://localhost:3000", // Websocket port
timeout: 5000
}))
.build();By default, VSLAMProcessingNode will create a single OpenVSLAM instance to handle the processing.
The framework is open source and is mainly developed by PhD Student Maxim Van de Wynckel as part of his research towards Hybrid Positioning and Implicit Human-Computer Interaction under the supervision of Prof. Dr. Beat Signer.
Use of OpenHPS, contributions and feedback is highly appreciated. Please read our contributing guidelines for more information.
Copyright (C) 2019-2022 Maxim Van de Wynckel & Vrije Universiteit Brussel
OpenHPS and the wrapper @openhps/openvslam is Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
stella-vslam (OpenVSLAM fork) is licensed under BSD and is not included in this repository.