COI Classifier

A desktop tool for rapidly scoring close-up photos with a single continuous COI value (0–100 %), plus collecting pixel-level liana / not-liana training samples. Same idea as a "Label box", but instead of drawing a box you drag a slider — and you can also click individual pixels to label them.

How it works

1. Launch — double-click run.bat, or run python coi_classifier.py.
2. Open a folder of close-up photos (File ▸ Open Folder, or the prompt on launch).
3. A random un-scored photo appears.
4. Zoom in to inspect detail, optionally label pixels (see below), then drag the slider (0–100 %) to set the COI.
5. Click Submit (or press Enter) — the score is recorded and the next random photo loads.

Moving across images

• Top slider — drag to scrub to any photo in the set.
• ◀ Prev / Next ▶ buttons, or PageUp / PageDown.
• The bar shows position / total and whether the current photo is ✓ scored or ○ unscored. Revisiting a scored photo restores its COI value and pixel labels. (Submit is still what records a score — just navigating past a photo doesn't score it.)

Zoom & pan

• Mouse wheel — zoom in / out, centred on the cursor
• Click + drag — pan around when zoomed in
• + / − / Reset — zoom buttons in the top bar (Reset = fit to window)
• Double-click — reset to fit

Pixel labelling (liana / not-liana)

• Tick “Label pixels” in the toolbar.
• Choose 🔴 Liana or 🔵 Not-liana (or press 1 / 2).
• Left-click a pixel to drop a labelled point — its (x, y) and RGB colour are sampled from the original image.
• Right-click a point to remove it; Undo point / Clear this photo buttons are in the toolbar.
• Markers stay pinned to the photo as you zoom and pan.

Shortcuts

• Enter — submit current photo
• ← / → — nudge the slider by 1 · Shift + ← / → — by 5
• 1 / 2 — select Liana / Not-liana
• Ctrl+S — Save project · Ctrl+Shift+S — Save As

Saving progress (projects)

Use the Save / Save As buttons (or the File menu) to store all progress — COI scores and pixel labels — in a .coiproj project file. You can close the app and reopen the project later (File ▸ Open Project) to carry on exactly where you left off. Once a project is saved, each Submit auto-saves to it. The app also warns about unsaved changes before closing.

Exporting data

From the File menu:

• Export COI CSV — one row per scored photo:

  filename	coi	timestamp
  IMG_0123.jpg	42	2026-06-24T14:05:31

• Export Pixel Labels CSV — one row per labelled pixel:

  filename	x	y	r	g	b	label
  IMG_0123.jpg	512	340	38	71	22	liana

The photo filename is the first column in both exports.

Requirements

• Python 3.8+
• Pillow — pip install pillow

Supported image types: jpg, jpeg, png, bmp, gif, tif, tiff, webp.

Command line

python coi_classifier.py                 # prompts for a folder
python coi_classifier.py path/to/photos  # open a folder directly
python coi_classifier.py session.coiproj # reopen a saved project