Question 1

What does the API return — a single hair type or a range?

Accepted Answer

The API returns a boundary range spanning two adjacent types. The mapping field provides a machine-readable string like "3a to 3b", and the term field gives the plain-language equivalent like "Loose to Medium Curls". This gradient output reflects real-world hair diversity — most hair sits between archetypes — and provides a richer signal for recommendation engines than a single hard label.

Question 2

What is the AI Hair Type Detection API used for?

Accepted Answer

The AI Hair Type Detection API analyzes a user's hair from three selfie photos and classifies it into one of nine hair type categories — from Type 1 (Straight) through Type 4C (Extremely Coily). It's designed for beauty brands, e-commerce platforms, salon tools, and beauty apps that want to deliver objective, AI-driven personalization without relying on customers to self-report their own hair type. Common use cases include product recommendation engines, virtual consultation flows, and first-party CRM enrichment.

Question 3

Why are three photos required instead of one?

Accepted Answer

Hair type cannot be reliably assessed from a single angle. Light reflection, product residue, and camera perspective can make wavy hair appear straight — or tight curls look loose — in one frame. The three-angle input (front, left 45°, right 45°) gives the model a full 180° view of the hair envelope, capturing both volume and curl tightness with far greater accuracy than any single-image approach. All three images must also be captured at the same pixel dimensions; mismatched sizes will return an error_mismatch_image_size error.

Question 4

What image requirements must be met to avoid errors?

Accepted Answer

Images must be JPG or PNG format, between 320px and 4096px in both dimensions, and under 10MB in file size. All three images must be the same pixel dimensions. The user's face must be fully visible — not cropped — and occupy roughly 50–80% of the frame width. For the front-facing shot, head tilt must stay within ±10° of straight. For the two side shots, the yaw angle must exceed 15° from center. The most common errors are error_insufficient_lighting (too dark), error_face_position_too_small (face too far from camera), and error_face_angle_invalid (side shots not angled enough). Hair should be worn down — not tied up — to allow accurate texture analysis

Question 5

What is the JS Camera Kit, and do I need it?

Accepted Answer

The JS Camera Kit is an optional but highly recommended component provided by Perfect Corp. It's a browser-embeddable photo capture flow that guides users through the three-angle selfie sequence in real time — providing live feedback on angle, framing, and lighting before the images are submitted. Without it, you'd need to build your own capture and validation UI. Using the Kit typically raises task success rates significantly by reducing invalid-photo submissions, which are the most common cause of failed detections. It integrates directly into mobile and desktop web experiences without requiring any native app development.

ビューティー

肌・顔・ボディ

ヘア & ひげ

ファッション

ジュエリー & ウォッチ

画像

動画

AI Hair Type Detection API

AI Hair Type Detection API

10 Hair Types - Full Spectrum Coverage

Powered by a 10-archetype curl detection model that maps every customer into one of nine precise categories — covering straight, wavy, curly, and coily hair in equal depth.

1: Straight

2A: Slight Wavy

2B: Medium Wavy

2C: Thick Wavy

3A: Loose Curls

3B: Medium Curls

3C: Tight Curls

4A: Kinky Soft

4B: Coily

4C: Extremely Coily