Multi-class eye shape identification with confidence scoring

The core model identifies 12 distinct eye shape categories, each with a confidence percentage. This allows businesses to set their own accuracy thresholds — for example, only applying automated recommendations when confidence exceeds 85%. The model handles variations in lighting and skin tone better than generic face classifiers because it was trained on a curated dataset of fashion and beauty images.

Business outcome: Product recommendations become anatomically relevant, increasing conversion rates on personalised suggestions.

RESTful API for real-time and batch processing

EyeShapeAI exposes a straightforward REST API that accepts image URLs or base64-encoded files. Response times average half a second per image, making the tool suitable for real-time virtual try-on flows. The API also supports batch endpoints for processing catalogue images in bulk, returning structured JSON with shape labels and confidence scores.

Business outcome: Engineering teams integrate eye shape classification without building custom computer vision models, saving months of development time.

Analytics dashboard for shape distribution insights

The web dashboard aggregates classification results to show shape distribution across your customer base or product catalogue. This data reveals which eye shapes are most common in your audience and which frame styles are being recommended most frequently. Exportable reports support offline analysis and campaign planning.

Business outcome: Merchandising teams make data-driven decisions about inventory mix and marketing focus based on real customer morphology data.

Client-side SDK for browser-based classification

A lightweight JavaScript SDK enables on-device eye shape classification directly in the browser, sending only the classification result — not the image — to the server. This approach reduces server load and addresses privacy concerns around uploading facial images to third-party servers. The SDK runs the model client-side using WebGL acceleration.

Business outcome: Privacy-conscious retailers offer personalised recommendations without storing or transmitting customer facial images to external servers.

Custom model fine-tuning for specialised use cases

For businesses with unique requirements — such as classifying children's eye shapes or analysing images from specific camera hardware — EyeShapeAI offers a fine-tuning service. Clients provide a labelled dataset, and the team retrains the model to improve accuracy on that specific distribution. This is priced separately from the standard API tier.

Business outcome: Companies with niche product lines achieve classification accuracy that off-the-shelf models cannot match, reducing false recommendations.

GDPR-compliant data handling with automatic deletion

The platform processes images with automatic deletion after classification unless explicitly stored by the client. No images are retained for model training without consent. The architecture supports data residency requirements for European and North American customers through configurable server regions.

Business outcome: Legal and compliance teams approve integration without custom data processing agreements, accelerating procurement cycles.

• Niche accuracy over general modelsEyeShapeAI's dedicated training on eye morphology outperforms general face classifiers that lump eye shape into broad, inaccurate categories.
• Real-time processing speedSub-second classification enables seamless integration into virtual try-on and recommendation flows without noticeable latency.
• Privacy-first architectureThe client-side SDK option eliminates the need to transmit facial images, addressing a major barrier to adoption in regulated markets.
• Actionable analyticsThe dashboard transforms raw classification data into merchandising insights that directly inform inventory and marketing decisions.

• Image quality dependencyPerformance degrades significantly on low-resolution, poorly lit, or angled images — the tool assumes professional-grade input.
• Limited to eye shape onlyBusinesses needing full facial analysis (skin tone, face shape, nose type) must integrate additional tools, increasing complexity.
• No offline modeThe SDK requires a web browser environment — there is no native mobile SDK or offline processing capability for in-store kiosks.
• Professional RealityEyeShapeAI solves one specific problem well, but fashion retailers expecting a complete visual AI solution will need to combine it with at least one other tool for full-face or body analysis.

Online eyewear retailer product matching

An eyewear ecommerce site integrates EyeShapeAI's API into its product detail page. When a customer uploads a selfie, the tool classifies their eye shape and surfaces frames that complement that morphology. The retailer reports a measurable reduction in return rates for online frame purchases. For broader fashion personalisation, this can complement tools like Vue.ai which handles full outfit recommendations.

Beauty brand virtual makeup try-on

A cosmetics brand uses the client-side SDK to classify eye shapes before applying virtual eyeshadow and eyeliner looks. The eye shape data adjusts the placement and style of digital makeup effects, creating a more realistic preview. This increases time spent in the virtual try-on tool and lifts add-to-cart rates for eye products.

Fashion catalogue enrichment at scale

A large marketplace runs EyeShapeAI's batch processing on its catalogue of model images. Each product photo receives eye shape metadata tags. The marketplace then uses these tags to power a 'find similar styles for your eye shape' filter, improving discovery for customers who know their eye type.

Marketing segmentation by eye morphology

A fashion brand's CRM team exports eye shape distribution data from the analytics dashboard. They discover that 40% of their audience has almond-shaped eyes. The team creates targeted email campaigns featuring models with that eye shape, resulting in higher click-through rates compared to generic campaigns.

Sign up for the Free plan on the EyeShapeAI website to receive an API key and access the developer documentation.

Upload a test set of 10–20 front-facing product or customer images via the API or dashboard to evaluate classification accuracy on your specific image types.

Review the analytics dashboard to understand the shape distribution in your test set and adjust your confidence threshold if needed.

Integrate the REST API into your product recommendation flow or use the client-side SDK for privacy-sensitive applications, then A/B test personalised recommendations against your current default.