wego-tray-tuner

Interactive tuning workspace for Wego tray inspection parameters, aligned with VisionRuntime detector logic.

wego-tray-tuner is a visual tuning tool built around detector logic synced from VisionRuntime. It focuses on parameter visibility, anchor editing, and overlay-based inspection review, so tuning can be done interactively instead of only through raw config files.

Overview

This project is intended for single-image detector tuning and review in a practical engineering workflow. It provides:

• visual parameter adjustment
• anchor placement and correction
• overlay-based inspection result review
• local working-config saveback after successful detection

Detector algorithm logic remains in the synced files under detect/, while this repository provides the UI shell and tuning workflow around that logic.

What this tool is for

wego-tray-tuner exists to make detector tuning more maintainable and reviewable.

Instead of editing inspection parameters only in YAML and re-running the full runtime, this tool lets you:

• load an image directly
• move anchors visually
• adjust parameters interactively
• run detection and inspect the returned overlay immediately
• save the current working state for continued tuning

This makes it useful as a focused tuning workspace for detector recipes that will later be used with VisionRuntime.

Relationship to VisionRuntime

This repository is not a separate detector implementation.

Its role is:

• UI and tuning workflow live here
• detector logic is synced from VisionRuntime under detect/
• detector recipes are loaded from config/wego_tray/

This structure helps keep tuning behavior aligned with the runtime behavior used in the main inspection project.

Highlights

• Interactive anchor placement for Wego tray inspection
• Visual parameter adjustment in a dedicated tuning UI
• Overlay-based inspection review after each detection run
• Neighbor-image browsing with Prev/Next for quick comparison
• Working-config persistence after successful tuning
• Synced detector logic for consistency with VisionRuntime

UI Preview

The UI provides interactive anchor placement, parameter editing, and overlay-based inspection review in a single workspace. It is designed to make detector tuning easier to validate visually before syncing recipes back into a runtime workflow.

Sync Workflow (Manual Copy)

Manually copy from VisionRuntime:

• VisionRuntime/config/wego_tray/*.yaml -> wego-tray-tuner/config/wego_tray/*.yaml
• VisionRuntime/detect/* -> wego-tray-tuner/detect/*
• VisionRuntime/utils/image_codec.py -> wego-tray-tuner/utils/image_codec.py
• optional samples -> wego-tray-tuner/data/images/*

Quick Start

pip install -r requirements.txt
python main.py

Packaging (PyInstaller onedir)

Build a folder-based package (onedir):

pip install pyinstaller
pyinstaller --noconfirm --clean --windowed --name wego-tray-tuner main.py

After build, copy the detector recipe folder to the generated output:

• source: config/wego_tray/*.yaml
• target: dist/wego-tray-tuner/config/wego_tray/*.yaml

Run:

• dist/wego-tray-tuner/wego-tray-tuner.exe

Notes:

• Keep the whole dist/wego-tray-tuner/ folder together (do not move only the exe).
• The packaged app reads/writes config/wego_tray/=@WORKING_+.yaml.
• The packaged app expects recipe files under config/wego_tray/.
• onedir output includes runtime dependencies under _internal/ (OpenCV / NumPy / Tk); this is normal.

UI Workflow

1. Click Open to load an image.
2. Drag 6 anchors (0/1/2 x L/R) and adjust parameters.
3. Click Detect to run inspection and update overlay.
4. Optional: use Prev/Next to browse images in the same folder.

After Detect, the displayed image is the detector overlay returned by detect/wego_tray.py.

Config Load/Save Behavior

Load priority

1. config/wego_tray/=@WORKING_+.yaml (working config)
2. first lexicographically sorted recipe under config/wego_tray/*.yaml, excluding _-prefixed files and =@WORKING_+.yaml

Save target after successful Detect

• config/wego_tray/=@WORKING_+.yaml

Write behavior

• If params were not changed, no file write occurs.

Project Layout

• main.py: thin application entrypoint
• app/__init__.py: app package export surface
• app/controller.py: app workflow controller
• ui/control_panel.py: right parameter panel
• ui/image_view.py: image canvas, overlays, anchors, nav hit areas
• core/slot_layout_utils.py: ROI grid generation helpers
• core/params.py: YAML load/merge/save
• core/detector.py: adapter to synced detector
• detect/: synced detector package from VisionRuntime
• config/: synced detector config
• data/: sample images

Docs

• docs/app_design.md: wego-tray-tuner app design and runtime flow
• docs/detect.md: detector reference and compatibility constraints
• docs/sync_contract.md: sync contract with VisionRuntime
• docs/ui_param_mapping.md: UI label to detector key mapping