bthek1/ros2_pi

How bthek1/ros2_pi works

ros2_pi is a ROS 2 monocular visual-odometry and dense-mapping system: a moving RGB camera supplies frames; tracking estimates camera pose while monocular inference estimates depth; mapping fuses those into a mesh viewed in a browser dashboard. The README explicitly notes there is no loop closure, pose graph, or relocalisation, so this is not a complete SLAM loop. The graph includes documented dataset replay and camera calibration capabilities. Some subsystem wiring is based on README and launch documentation because those implementation files were not sampled.

Main entry points: pimesh.launch.py ROS bringup · odometry_node.cpp Pose estimation · fusion_node.cpp TSDF fusion · depth_node.cpp Depth node

Components 22

Capture and bringup

Camera capture src/pimesh_camera/src/camera_node.cpp
ROS bringup src/pimesh_bringup/launch/pimesh.launch.py
Camera calibration tools/calib/calib_solve.py

Visual tracking

Frame decode src/pimesh_frontend/src/decode_node.cpp
ORB keypoints src/pimesh_frontend/src/keypoint_node.cpp
Pose estimation src/pimesh_frontend/src/odometry_node.cpp
Keyframe store src/pimesh_frontend/src/keyframe_store.cpp

Depth inference

Depth node src/pimesh_depth/src/depth_node.cpp
Depth engine src/pimesh_depth/src/depth_engine_ort.cpp

Dense mapping

TSDF fusion src/pimesh_mapping/src/fusion_node.cpp
TSDF volume src/pimesh_mapping/src/tsdf_volume.cpp
Shared volume src/pimesh_mapping/src/shared_volume.cpp
Mesh extraction src/pimesh_mapping/src/mesh_node.cpp
Mesh processing src/pimesh_mapping/src/marching_cubes.cpp

Replay and dashboard

Dataset replay src/pimesh_dataset/src/dataset_node.cpp
Dataset reader src/pimesh_dataset/src/dataset_reader.cpp
Dashboard node src/pimesh_dashboard/src/dashboard_node.cpp
Web server src/pimesh_dashboard/src/web_server.cpp
WebSocket src/pimesh_dashboard/src/websocket.cpp

Other components

RGB camera
Browser viewer src/pimesh_dashboard/web/app.js
ONNX model

Connections 32

RGB camera → Camera capture: captures frames
Camera capture → Frame decode: publishes JPEG
Camera capture → Pose estimation: publishes intrinsics
Camera capture → TSDF fusion: publishes intrinsics
ROS bringup → Frame decode: loads component
ROS bringup → ORB keypoints: loads component
ROS bringup → Pose estimation: loads component
ROS bringup → Depth node: loads component
ROS bringup → TSDF fusion: loads component
ROS bringup → Mesh extraction: loads component
Frame decode → ORB keypoints: publishes frames
Frame decode → Depth node: publishes frames
ORB keypoints → Pose estimation: publishes tracks
Pose estimation → Keyframe store: updates keyframes
Depth node → Depth engine: runs inference
ONNX model → Depth engine: provides weights
Depth node → TSDF fusion: publishes depth and RGB
Pose estimation → TSDF fusion: publishes poses
TSDF fusion → Shared volume: accesses volume
Shared volume → TSDF volume: shares TSDF
Mesh extraction → Shared volume: reads volume
Mesh extraction → Mesh processing: extracts surface
Mesh extraction → Dashboard node: publishes mesh
Depth node → Dashboard node: publishes status and images
Pose estimation → Dashboard node: publishes pose and stats
TSDF fusion → Dashboard node: publishes stats
Dataset replay → Dataset reader: reads recordings
Dataset reader → Frame decode: replays frames
Dashboard node → Web server: serves dashboard
Web server → WebSocket: dispatches messages
WebSocket → Browser viewer: streams updates
Browser viewer → Web server: requests actions

More C++ diagrams: uv-software/pcanbasic-wrapper · amd/vitis-ai · aussiemakergeek/esp32-multi-room-intercom · axdhill/ait-qkd · charliebrej/bombe

Drawn by GitDiagram on September 28, 2026 from the repository's file tree, README and source files. Read as Markdown ·