Image Occlusion

What image occlusion is

Image occlusion hides specific regions of a labeled diagram behind colored masks. Each mask becomes a flashcard question: you see the diagram with one or more regions covered and must recall what is beneath each mask. Revealing the mask confirms your answer.

The technique is particularly effective for subjects where spatial context is part of the knowledge: anatomy (muscle and bone labels in position), geography (country and capital locations on a map), chemistry (functional groups in a molecular structure), and engineering diagrams (component labels in a schematic).

For the cognitive science behind why image occlusion produces stronger memories than text-only flashcards, see Dual Coding in the learning science section.

Getting to the tool

The Image Occlusion tool is at neurako.com/tools/image-occlusion. No account or login is required to use the tool or export cards.

Uploading an image

Drag and drop or click to upload a .png, .jpg, or .webp file. Images are automatically downscaled to a maximum of 2048px on the longest edge for display and export. All mask coordinates are stored as normalized ratios (0–1 relative to the original image dimensions), so downscaling never shifts or invalidates your masks.

Accepted file types: PNG, JPEG, WebP. Maximum recommended file size: 20 MB before downscaling.

Drawing masks manually

After uploading, select a drawing tool from the toolbar:

Rectangle — Click and drag to draw a rectangular mask. Best for text labels, callout lines, and most diagram annotations.

Ellipse — Click and drag to draw an elliptical mask. Useful for circular structures, cell diagrams, or labels inside rounded shapes.

Polygon — Click to place points; double-click or press Enter to close the shape (minimum 3 points). Use for irregular boundaries, country outlines, or non-rectangular anatomical regions.

Editing masks

After drawing, masks can be:

• Moved by dragging
• Resized using the handles at the corners and edges
• Labeled by clicking the label field in the Mask List sidebar — the label is used for the "Back Extra" field and is shown in the mask list for organization
• Grouped with other masks by selecting multiple masks and pressing G — grouped masks share the same cloze ordinal and reveal together in Hide All / Guess One mode

Keyboard shortcuts

Undo / redo

The editor maintains a full history of mask operations. Use Cmd/Ctrl+Z to undo any number of steps, and Shift+Cmd/Ctrl+Z to redo. History is preserved until you upload a new image or reset the editor.

AI auto-detection (Gemini BYOK)

The AI panel provides automatic mask generation using Google Gemini's vision capabilities. Gemini analyzes your image and returns bounding boxes and labels for the regions it identifies as likely study targets — text labels, callouts, named structures.

Using AI auto-detect

1. Open the AI panel in the right sidebar and enter your Gemini API key.

2. Optionally enter a focus hint to guide detection — for example, "muscle names" or "country capitals" — to narrow Gemini's attention to specific label types.

3. Click "Auto-detect regions." Detection typically takes 8–12 seconds.

4. Review the generated masks on the canvas. Accept, adjust, or delete as needed before exporting.

AI-detected masks are identical to manually drawn masks — they appear on the canvas, can be edited, labeled, grouped, and exported in the same way. AI detection augments your work rather than replacing it: if you already have manual masks, AI masks are merged in without overwriting existing ones (duplicate regions are de-duplicated by overlap detection).

Supported providers for AI detection: Google Gemini (AI Studio key), and Google Gemini via OpenRouter. Vision-capable Gemini models (Gemini 2.5 Flash and later) are recommended for best accuracy.

Occlusion modes

Two modes control how masks behave across the cards generated from a single image.

Hide All, Guess One

All masks are hidden on the front of every card. Each card in the deck reveals exactly one mask on the back — all others remain hidden. A single image with 10 masks produces 10 separate cards.

This is the recommended default for most subjects. Hiding all labels prevents using visible neighbor labels as context clues, producing stronger recall.

Hide One, Guess One

Only the target mask is hidden on the front. All other masks show their labels. This mode provides more context and is appropriate for:

• Early-phase learning of a complex diagram, before you have a mental map of the region
• Diagrams where the relative positions of other labeled structures are themselves informative
• Small diagrams where hiding all labels would leave insufficient context

Select the mode in the right sidebar before exporting. Mode is stored per card set and encoded in the exported Anki note type.

Card metadata

The right sidebar contains fields that apply to all cards in the current image's set:

Header — shown at the top of each card. Use this for the subject or section name (e.g., "Shoulder — Rotator Cuff Muscles").

Back Extra — additional content shown on the answer side. Use for clinical correlations, mnemonics, origin/insertion/action for anatomy cards, or any context that adds value beyond the label itself.

Comments — internal notes not shown on the card face. Useful for your own reference or attribution.

Tags — Anki-compatible tags applied to all cards in the export (e.g., anatomy::upper-limb, USMLE::Step1).

Previewing cards

The Preview panel shows the front and back of a representative card using the current masks, mode, and metadata. Use this to verify the mask coverage and card appearance before exporting.

Exporting

1. Click the Export button in the sidebar or toolbar once you have at least one mask drawn.

2. Choose your export format.
3. The file downloads immediately to your device.

Export formats

Anki Package (.apkg) — Recommended. Generates a native Anki Image Occlusion deck importable by Anki 23.10+, AnkiMobile, AnkiDroid, and AnkiWeb. One import creates all cards with correct cloze ordinals, embedded image, and metadata. Uses Anki's native IO note type (Occlusion, Image, Header, Back Extra, Comments fields) — not the legacy Image Occlusion Enhanced add-on format.

JSON — Portable bundle containing all card data and embedded images. Can be reimported into the Neurako Image Occlusion tool for further editing.

CSV — Flat text export with columns for image filename, occlusion field text, header, back extra, comments, and tags. For use with other flashcard platforms that accept CSV import.

Multi-image decks

You can add multiple images to a single export session. After uploading and masking the first image, use the image rail at the top of the editor to add additional images. Each image maintains its own independent set of masks, mode, and metadata. All images are included in a single .apkg export.

Using with Anki

Importing the .apkg

1. Open Anki desktop (23.10+)
2. Go to File → Import and select the downloaded .apkg file
3. Choose the target deck or create a new one
4. Click Import

All cards will appear in your selected deck with the image and masks correctly configured. Native IO cards display with Anki's built-in Image Occlusion renderer — no add-on required.

Editing cards in Anki

After importing, individual masks can be adjusted in Anki's card editor. Open any IO card in the editor and use Anki's built-in mask editor (click the image) to move, resize, or add masks.

FSRS scheduling

Anki 23.10+ includes built-in FSRS support. Enable it in Tools → Preferences → Review → FSRS to use the adaptive scheduling algorithm with your image occlusion cards.

Privacy and data handling

The Image Occlusion tool in apps/tools (neurako.com/tools) is a fully client-side tool:

• Images are loaded and processed entirely in your browser
• No image data is transmitted to Neurako's servers at any point
• If AI auto-detection is used, the image is sent directly from your browser to Google's Gemini API using your own API key — Neurako's infrastructure is not involved in that request
• Your Gemini API key is stored only in your browser session and is never persisted or transmitted to Neurako

Subject-specific guidance

For detailed workflows by subject area, see:

Frequently asked questions

Related journal guide

For a narrative introduction to image occlusion — the research behind it, best practices by subject, and a detailed walkthrough of the AI workflow — read Image Occlusion Flashcards: The Definitive Guide on the Neurako Journal.

Related reading

AI Flashcard Generation

Generate flashcard drafts from notes, text, images, audio, or uploaded files using Neurako's AI creation tools.

Capture from Images

Photograph a textbook page, slide, whiteboard, or screenshot and Neurako will extract flashcard drafts using OCR and AI.

Memorizing Anatomy

How medical and allied-health students retain hundreds of named structures — concept maps, mnemonics, 3D visualization, spaced repetition, and the research on what actually works.