SLAM@HOME

An implementation of a real-time SLAM system over a local Wi-Fi network. This project was initially started for my self-learning purpose so the implementation is not production ready and may include performance issues and/or edge cases but I still believe this can be a code example for something like a DIY 3D scanner project or a DIY 3D survelliance system. Hope you will like it.

Architecture

This diagram illustrates the overall architecture of SLAM@HOME.

Implementation Notes

1. React/Frontend

   All components (except the markdown parser and social buttons) are built using vanilla React.

2. Express/Signaling

   This is the possibly simplest implementation of a signaling server for WebRTC. It only provides the minimal set of functionalities required to exchange Session Description Protocols for establishing peer connections.

3. iOS/Camera

   A monocular depth estimater/sampler implementation for iOS. The estimated and/or sampled depth data is streamed via a peer connection to the frontend. The WebRTC SDK is GoogleWebRTC and the depth estimation is based on a machine learning model for now [4/17/2022] but this will be replaced with LiDAR camera due to the incompatible design of the SDK with AVFoundation causing performance issues. App Transport Security restrictions are disabled since the system is supposed to be deployed only on your local network.

4. MLModels

   A collection of machine learning models which is used for iOS/Camera. This directory may be deprecated someday due to the reasons mentioned above.

5. Wgpu/Core

   The SLAM core engine for reconstructing 3D models from the video stream. The engine will be implemented in Rust.

The following is a checklist of features and their progress:

• Documentation
  • README
• DevOps
  • Dockerfile
  • Kubernetes
• React/Frontend
  • WebRTC
• Express/Signaling
  • WebRTC Signaling
  • Device Detection
• iOS/Camera
  • ML based Depth Sampler
  • CPU Monitor
  • LiDAR Camera over DataChannel
• MLModels
  • Pydnet
• Wgpu/Core
  • SLAM engine
• Android
• Video Server
  • Recording