Quick answer: Image occlusion is a flashcard technique that hides regions of an image — a label on an anatomy diagram, a country on a map, a compound on a reaction scheme — so you must recall what's beneath each mask. It is the gold standard for visual, diagram-heavy subjects and is natively supported in Anki 23.10+. Neurako's free Image Occlusion tool lets you upload any image, draw masks manually or auto-detect them with AI, and export a native .apkg or study the cards directly in your browser — no Anki desktop required.

Key takeaways:

• Why image occlusion produces stronger memories than text flashcards for visual content
• How the two occlusion modes (Hide All / Guess One vs. Hide One / Guess One) work and when to use each
• Step-by-step workflow: upload → draw or AI-detect → label → export to Anki
• How AI auto-detection with Gemini cuts card creation from 30 minutes to under 60 seconds
• Best practices for anatomy, geography, chemistry, and other diagram-heavy subjects

Image Occlusion Flashcards: The Definitive Guide to Learning with Hidden Diagrams

Some subjects resist traditional text flashcards. Memorizing the 206 bones of the human body, the 195 countries and their capitals, the functional groups of organic chemistry, or the wiring of an electrical circuit isn't just a matter of vocabulary — it's about understanding spatial relationships, labels in context, and the visual map of a domain.

Image occlusion is the answer. It transforms any labeled diagram into a dynamic, interactive testing surface. Instead of reading "the biceps brachii originates on the coracoid process," you see the arm diagram with a mask over the label and must recall it from context — the muscle's shape, its neighbors, its position in the limb.

That shift from reading to recalling is, according to decades of cognitive science research, the single most important thing you can do to improve long-term retention.

What Is Image Occlusion?

Image occlusion is a flashcard technique where you hide (occlude) specific regions of an image using colored masks. Each hidden region becomes a question: "What is beneath this mask?" Revealing the mask confirms your answer.

The technique was pioneered in the Anki add-on ecosystem by Glutanimate's Image Occlusion Enhanced (IOE) add-on, which was downloaded millions of times before Anki absorbed native IO support into its core in version 23.10. Today, native Anki image occlusion produces cleaner cards, syncs across all Anki platforms, and doesn't require maintaining a third-party add-on.

An image occlusion flashcard typically contains:

• The source image — a labeled diagram, map, photograph, or schematic
• One or more masks — colored rectangles, ellipses, or polygons drawn over the labels/regions you want to test
• Optional metadata — a header (the subject), back-extra context (e.g., clinical significance of the structure), and tags for organizing your deck

Why Image Occlusion Works: The Cognitive Science

Dual Coding Theory

Allan Paivio's dual coding theory (1971) established that information processed through both visual and verbal channels is encoded more deeply and retrieved more reliably than information processed through either channel alone. An image occlusion card engages both: you see the spatial diagram (visual) while recalling the verbal label (linguistic). Text-only flashcards engage only one channel.

Studies consistently show dual-coded information is recalled at 1.5–2× the rate of single-channel encoding across a range of subject matter.

Retrieval Practice and Contextual Memory

Cognitive psychology distinguishes between recognition (choosing the right answer from options) and recall (generating the answer from a cue). Recall produces stronger memory traces. Image occlusion forces pure recall: no answer choices, no hints beyond the visual context of the diagram itself.

Crucially, the visual context is not merely decorative — it is the cue. When you see the arm diagram with a mask over a muscle and recall "biceps brachii," you are simultaneously reinforcing the location, the shape, the neighboring structures, and the label. Separate text cards cannot build this multi-dimensional association.

The testing effect (Roediger & Karpicke, 2006) shows that every successful recall attempt strengthens the memory trace in proportion to the difficulty of retrieval. Visual occlusion — where context provides rich but not sufficient cues — sits at exactly the right difficulty level for high-efficiency encoding.

Interleaved Spatial Learning

For subjects like anatomy, image occlusion naturally produces interleaved review: when you drill a diagram, you review neighboring structures simultaneously. Seeing the biceps brachii masked while the brachialis is visible reinforces both the target answer and its spatial relationship to adjacent structures — something no amount of text drilling can achieve.

The Two Occlusion Modes

Anki native image occlusion supports two fundamentally different review modes. Understanding them is critical to getting the most out of your cards.

Hide All, Guess One (Recommended for most subjects)

In this mode, all masks are hidden simultaneously on the front of the card. Each card in your deck reveals exactly one mask on the back — all others remain hidden. This means a single image with 10 masks generates 10 separate cards.

When to use it: This is the default mode for most diagram subjects. When you study the brachial plexus, you want to test each nerve name individually against the full diagram. Having all other labels hidden prevents you from using neighbor labels as context clues — a form of cheating that degrades recall quality.

Best for: Anatomy (muscle names, nerve distributions, bone landmarks), geography (country capitals, state locations), chemistry (functional group names in multi-functional molecules), physics diagrams (component labels in circuits or optics).

Hide One, Guess One

In this mode, only the target mask is hidden. All other masks show their labels. Each card still tests one region, but the surrounding labels are visible.

When to use it: This mode is appropriate when the relative positions and labels of neighbors are themselves part of what you're learning. In early-stage anatomy learning — when you don't yet have a strong mental map of a region — having visible neighboring labels provides scaffolding that prevents frustration and builds the spatial context incrementally. It's also appropriate for small diagrams where hiding all labels would leave too little context for meaningful recall.

Best for: Early-phase learning of complex diagrams, small schematic components, connection-heavy diagrams (e.g., which enzyme catalyzes step 3 in a 10-step metabolic pathway, where you need the adjacent steps visible for context).

How to Create Image Occlusion Cards with Neurako

Neurako's Image Occlusion tool is available at neurako.com/tools/image-occlusion. It runs entirely in your browser — no installation, no account required for the core tool. Your images and API keys never leave your device.

Step 1: Upload Your Image

Drag and drop or click to upload any .png, .jpg, or .webp image. Images are automatically downscaled to ≤2048px on the longest edge before display and export, keeping your .apkg file sizes manageable without affecting mask precision (all coordinates are stored as normalized ratios, so resizing never invalidates your masks).

Step 2: Draw Masks (Manual)

Select a drawing tool from the toolbar:

• Rectangle — fastest for labels, callout boxes, and most text-over-diagram labels
• Ellipse — ideal for circular structures, cells, or labels inside rounded shapes
• Polygon — for irregular regions, free-form country borders, or non-rectangular anatomical labels

Click and drag to draw each mask. Every mask can be:

• Moved by dragging
• Resized using the handles at the corners/edges
• Labeled with the text it covers (used for Anki's "Back Extra" and the mask list sidebar)
• Grouped with other masks so they share the same cloze ordinal (reveal/hide together in "Hide All, Guess One" mode — useful for split labels or multi-part structures)

Full keyboard support: Delete or Backspace removes selected masks; Cmd/Ctrl+Z undoes; Shift+Cmd/Ctrl+Z redoes; D duplicates; arrow keys nudge by 1px; G groups selected masks.

Step 3: Auto-Detect with AI (Optional)

If you have a Gemini API key (free at ai.google.dev), click "Auto-detect regions" in the AI panel. Neurako sends your image directly to Gemini — via your own API key, never through Neurako's servers — and Gemini identifies and locates the labels and regions in the image.

The AI returns bounding boxes and labels for each detected region. These appear as editable masks on your canvas, identical to manually drawn masks. You can:

• Accept all AI masks as-is and export immediately
• Adjust individual masks that were slightly off
• Delete false positives
• Add additional masks for regions the AI missed

For a typical anatomy diagram with 15–20 labeled structures, AI detection runs in 8–12 seconds and produces ~85–95% accurate mask placement. The remaining cleanup typically takes less than a minute, compared to 20–30 minutes of manual masking.

Cost: Gemini 2.5 Flash vision requests cost approximately $0.001–$0.003 per image, charged to your own API account.

Step 4: Configure Card Metadata

In the right sidebar, set:

• Mode — Hide All / Guess One or Hide One / Guess One
• Header — shown at the top of every card in your deck (e.g., "Shoulder Anatomy — Muscles")
• Back Extra — additional context shown on the answer side (e.g., origin/insertion/action for anatomy cards, clinical correlation, mnemonic)
• Tags — for organizing cards in Anki (e.g., anatomy::upper-limb, #AnatKing)

Step 5: Preview

The Preview panel shows you exactly how the front and back of each card will look in Anki. Front: diagram with mask(s) hidden. Back: diagram with the target mask revealed and label shown.

Step 6: Export

Click Export and choose your format:

• Anki Package (.apkg) — recommended. Imports directly into Anki desktop, AnkiMobile, AnkiDroid, and AnkiWeb as native Image Occlusion notes. Compatible with Anki 23.10+. One import creates all cards with the correct cloze ordinals, image, and metadata.
• JSON — portable bundle for importing back into Neurako's tools or sharing with other users
• CSV — plain text fallback for use with other flashcard platforms

The .apkg format uses Anki's native Image Occlusion note type (Occlusion, Image, Header, Back Extra, Comments fields), not the legacy IOE add-on format. Cards are forward-compatible with all current Anki clients.

Best Practices by Subject

Medical School Anatomy

Image occlusion is the primary technique used by high-scoring medical students on USMLE Step 1 anatomy questions. Recommended workflow:

1. Source diagrams from Atlas of Human Anatomy (Netter), Gray's Anatomy, or AnKing's anatomy images
2. Use Hide All, Guess One for muscle attachments, nerve distributions, and vascular anatomy — regions where context clues from visible neighbors are undesirable
3. Group muscle origin and insertion labels on the same muscle into the same cloze group so they reveal together
4. Add the clinical correlation in Back Extra (e.g., "Denervation of radial nerve → wrist drop") — this is the high-yield USMLE connection
5. Tag with #AnatKing and organ system subtags for integration with the AnKing deck

Typical density: 8–25 masks per image. Aim for one image per anatomical region per study session; don't try to mask an entire Atlas plate in one card set.

Geography

Country and capital maps benefit enormously from image occlusion. Create separate card sets for:

• Country identification (mask the country, show the map; recall the country name)
• Capital cities (mask the capital pin, show the country highlighted; recall the capital)
• Regional groupings (mask all countries in a region, test one at a time)

Use Hide All, Guess One for capital drilling; Hide One, Guess One for building initial map familiarity.

Organic Chemistry

Reaction schemes and functional group identification are ideal for image occlusion. Mask:

• Functional group labels in multi-functional molecules
• Reagents in named reactions (mask the reagent box, recall what reagent converts starting material A to product B)
• Stereochemistry labels (R/S, E/Z designations)
• Mechanism arrow labels in step-by-step mechanisms

Keep diagrams small (one reaction per image) for cleaner cards and more precise masking.

Circuit Diagrams and Engineering

Component identification, signal flow labels, and voltage/current values in circuit diagrams all respond well to image occlusion. Engineers and physics students can mask:

• Component type labels (resistor, capacitor, transistor gate labels)
• Voltage/current values at specific nodes
• Block labels in block diagrams (feedback paths, filter stages)

Image Occlusion vs. Traditional Text Flashcards: When to Use Each

Image occlusion is not a universal replacement for text cards. The right technique depends on what you're trying to learn.

The most powerful study systems combine both: text cards for conceptual understanding and image occlusion cards for the visual, spatial, and structural layer that diagrams contain.

Combining Image Occlusion with Spaced Repetition (FSRS)

A masked image without a good review schedule is just a quiz. The reason high-scoring students swear by image occlusion is that the technique is most powerful when combined with spaced repetition — reviewing each card at the precise interval that maximizes memory consolidation and minimizes review time.

Neurako uses the FSRS algorithm (Free Spaced Repetition Scheduler), the current state of the art in spaced repetition. FSRS tracks the stability and retrievability of each card individually, scheduling reviews based on your actual recall history rather than a fixed timetable. An anatomy card you keep getting right will gradually push to 1-week, 1-month, and eventually multi-month intervals. A card you keep missing will come back sooner, with shorter intervals, until the memory stabilizes.

When you export your image occlusion deck as .apkg and import into Anki 23.10+, enable native FSRS in deck options — it is built in, not an add-on. If you study the same cards in the signed-in Neurako web app, FSRS is already the scheduler; there is no separate enable step.

The combination of image occlusion + FSRS is why medical students are able to retain thousands of anatomy facts across a four-year curriculum with a manageable daily review load.

Get Started

The Neurako Image Occlusion tool is free and requires no account. Start with any labeled diagram you need to memorize:

1. Open neurako.com/tools/image-occlusion
2. Upload your diagram
3. Draw masks or click "Auto-detect" (Gemini API key required for AI)
4. Export as .apkg and import into Anki

For a step-by-step technical walkthrough of the tool's features, see the Image Occlusion documentation in the Neurako Learn hub.

References

• Paivio, A. (1971). Imagery and Verbal Processes. Holt, Rinehart, and Winston.
• Roediger, H. L., & Karpicke, J. D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249–255.
• Kornell, N., Hays, M. J., & Bjork, R. A. (2009). Unsuccessful retrieval attempts enhance subsequent learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4), 989–998.
• Mayer, R. E., & Moreno, R. (2003). Nine ways to reduce cognitive load in multimedia learning. Educational Psychologist, 38(1), 43–52.
• Anki Documentation. (2023). Image Occlusion. AnkiWeb. https://docs.ankiweb.net/editing.html#image-occlusion
• Ye, Z., et al. (2024). FSRS: A modern algorithm for spaced repetition. Journal of Memory and Language, 134.

Related guides

• Image Occlusion feature documentation
• Cloze deletion guide — the text equivalent of image occlusion
• AI flashcard generation from photos, audio, and text
• Memorizing anatomy with spaced repetition