📘 Introduction

• Sense Core (s6)
  • Features
  • Installation
  • Quick Start
  • Common Commands
  • Tracking Runtime
  • Unit Testing
    • Test Design Guidelines
    • Running Tests
  • Profiler
• System Architecture Overview
  • Overview
  • Execution Flow
  • Context Generators
    • ContextGenerator
    • RemoteGSTContextGenerator
    • LocalGSTContextGenerator
    • DatasetContextGenerator
  • Pipeline Architecture
    • Concrete Pipelines
  • Runtime Modes
    • Headless mode
    • UI mode
    • Uplink mode
  • Retired Components
• GUI Architecture & Interaction
  • 1. High-Level Layout
  • 2. Modes & Initialization
  • 3. Canvas & Visual Binding
    • 3.1 Views
  • 4. Update Loop & Context Contract
  • 5. Control Panel Widgets
    • 5.1 PlaybackControlWidget (Replay mode)
    • 5.2 CaptureToolbar (Live mode)
    • 5.3 DatasetSelectorWidget
    • 5.4 Exposure / Contrast Controls
    • 5.5 RoboticControlPanel (Live mode)
    • 5.6 Logger
  • 6. Commands Emitted by the GUI
  • 7. Data Flow
    • 7.1 Event Flow (User → Backend)
    • 7.2 Telemetry Flow (Backend → GUI)
  • 8. Timing & Performance Notes
  • 9. Extensibility Guidelines
  • 10. Known Assumptions
• command_queue Control Channel
  • 1. Purpose and Shape
  • 2. ContextGenerator Usage
  • 3. Commands and Consumers
  • 4. track.py Integration
  • 5. GUI Integration
• Pipeline Configuration
  • How It Loads
  • Schema Overview
  • Using Config in Stages
  • Extending the Schema (add a new parameter)
    • Design Tips
  • Changelog Highlights (recent)
  • Troubleshooting
• Dataset Storage and R2 Utilities
  • Concepts
  • Motivation
  • Credentials and Defaults
  • Setup
  • CLI: s6 r2 Utilities
    • List
    • Download
    • Upload
    • Delete
  • CLI: s6 dataset Master Command
    • List
    • Upload
    • Download
    • Delete
  • R2Client API
    • Design Overview
    • Construction
    • Listing
    • Download
    • Upload
    • Delete
    • Concurrency and Progress
  • Safety Semantics
  • Examples
    • Multi‑User Workflow: Record → Upload → Replay
    • End-to-end roundtrip for a dataset named diverse_2
  • Troubleshooting
• Legacy Streamer Retirement (TCP/HTTP)
  • What was removed
  • Current s6 track input sources
  • Impacted commands/scripts (retired for now)
  • Restored with gst path
  • Partially impacted utilities
• Applications (s6 CLI)
  • camera_preview
  • cog/keypoint — Keypoint Training, Preview, And ONNX Export
    • Common Usage
    • Command Modes
    • Dataset Config
    • Model And Training Behavior
    • Logging And Checkpoints
  • Train explicitly with bf16 autocast on CUDA
  • Train for 50 epochs and export the checkpoint from that same run to the
  • auto-generated assets/models/_… path
  • Export a fixed batch-2 ONNX model from the latest saved checkpoint
  • Override the output path explicitly when needed
    • ONNX Export
    • How it works
    • Key flags
  • data/collect — Retired
  • data/filter — Interactive dataset browser and filter
    • Usage
    • Controls
    • How it works
  • dataset — Manage datasets (local and R2)
    • Usage (selected)
    • How it works
    • Common flags
    • Subcommands
    • Examples
  • id — Retired
  • perf-stats — Compare profiling runs
    • Usage
    • How it works
  • Placeholder
  • robotic/remote-control — Qt GUI for remote control
    • Usage
    • How it works
    • Notes
  • robotic/server — Robotics REST API server
    • Usage
    • Key endpoints (prefix /robotic)
    • How it works
  • Blender Simulation Renderer (s6 sim render)
    • Why use it
    • Quick start
    • Output layout
    • Replay in s6 track
    • Notes on calibration
    • Object Logging
    • See also
    • Install Blender Python Packages
    • Calibration CLI (s6.app.sim.calib)
    • ChArUco Utilities (s6.app.sim.charuco_detect)
    • Troubleshooting
  • stream — Retired
  • test — Run the project’s unit tests
    • Usage
    • How it works
  • track — Run the tracking stack
    • Usage (selected)
    • How it works
    • Common flags
    • Retirement note
    • Gst Source Config
  • uplink — Push mock telemetry into a visualizer
    • Usage
    • How it works
• Recipes
  • Recipe: Dataset Capture + Dev Loop
    • 0) Prerequisites
    • 1) Record from the gst source
    • 2) Develop and test the pipeline on the dataset
    • 3) Optional: augment or refine the dataset
    • Troubleshooting
  • Recipe: Model Refine
    • 0. Prerequisites
    • 1. Record while running inference
    • 2. Filter out wrong predictions
    • 3. Prepare training config
    • 4. Preview data
    • 5. Train a finer model
    • 6. Put the model back into the pipeline
  • Recipe: Pipeline Chrome Trace
    • What you’ll get
    • 0) Prerequisites
    • 1) Run track with logging enabled
    • 2) Locate the trace files
    • 3) Inspect the timeline
    • 4) Optional: summarize stats
    • Troubleshooting
• s6.app
  • s6.app.pipeline
    • s6.app.pipeline.base
    • s6.app.pipeline.t1
    • s6.app.pipeline.v1
    • s6.app.pipeline.v2
  • s6.app.r2
    • s6.app.r2._common
    • s6.app.r2.delete
    • s6.app.r2.download
    • s6.app.r2.list
    • s6.app.r2.test
    • s6.app.r2.upload
  • s6.app.robotic
    • s6.app.robotic.control
    • s6.app.robotic.server
  • s6.app.sim
    • s6.app.sim.calib
    • s6.app.sim.install_package
    • s6.app.sim.render
    • s6.app.sim.render_animation
  • s6.app._common
  • s6.app._contextgenerators
  • s6.app._gui
  • s6.app._pipeline
  • s6.app._preview
  • s6.app._vispy
  • s6.app.camera_preview
  • s6.app.dataset
    • Defaults
  • s6.app.ds
  • s6.app.id
  • s6.app.keypoint_runtime
  • s6.app.main
  • s6.app.monitor
  • s6.app.replay
  • s6.app.test
  • s6.app.track
  • s6.app.uplink
• s6.app.data
  • s6.app.data.collect
  • s6.app.data.convert
  • s6.app.data.filter
  • s6.app.data.instrument_render_v3
  • s6.app.data.render
  • s6.app.data.visualize
• s6.app.cog
  • s6.app.cog.augmented_dataset
  • s6.app.cog.benchmark
  • s6.app.cog.dataloader
  • s6.app.cog.keypoint
  • s6.app.cog.trt_runtime_experiment
  • s6.app.cog.yolo
• s6.vision
  • s6.vision.camera
    • Key concepts
  • s6.vision.detectors
  • s6.vision.drawing
  • s6.vision.preprocess
  • s6.vision.solver
  • s6.vision.tracking
  • s6.vision.trajectory
    • Key concepts
• s6.schema
  • s6.schema.calibration
  • s6.schema.connection
  • s6.schema.data
  • s6.schema.gstreamer
  • s6.schema.pipeline_config
  • s6.schema.pipeline_settings
  • s6.schema.platform
  • s6.schema.primitives
    • Coordinate conventions
  • s6.schema.robotic
    • Key concepts
  • s6.schema.trace
    • Key concepts
• s6.robotic
  • s6.robotic.client
  • s6.robotic.service
  • s6.robotic.stepper
• s6.rendering
  • s6.rendering.visuals
• s6.nn
  • s6.nn.data
    • s6.nn.data.aug
    • s6.nn.data.augmentation
    • s6.nn.data.dataset
  • s6.nn.ops
    • s6.nn.ops.layers
  • s6.nn.keypoint_detector_v2
  • s6.nn.keypoints
  • s6.nn.runtime
  • s6.nn.trt_runtime
  • s6.nn.utils
• s6.utils
  • s6.utils.blender
  • s6.utils.bounding_box_selector
  • s6.utils.calibration
  • s6.utils.camera_capture
  • s6.utils.camera_identifier
  • s6.utils.camera_identifier_aruco
  • s6.utils.camera_identifier_bottom
  • s6.utils.camera_identifier_interface
  • s6.utils.cleartext
  • s6.utils.datapipe
  • s6.utils.devices
  • s6.utils.filters
  • s6.utils.infra
  • s6.utils.logging_utils
  • s6.utils.os
  • s6.utils.path_patterns
  • s6.utils.paths
  • s6.utils.profiler
  • s6.utils.r2
    • Key concepts
    • Motivation
    • Design
  • s6.utils.stage_decorator
  • s6.utils.sync_controller
    • Endpoints
  • s6.utils.tensor_registry
  • s6.utils.waypoint_utils
• Docstring Style Guide (for Sphinx Autodoc)
  • Quick Rules
  • Module Docstrings
  • Class Docstrings
  • Function and Method Docstrings
  • Pydantic Models and Data Containers
  • Special Cases
  • Cross-References
  • Notes and Examples
  • Style Details
  • Checklist
  • Building the Docs