Body Measurement Bundles Explained: TORSO, HAND_ARM, LEGS_FEET, HEAD_FACE

If you’re integrating a body measurement API, one of the first design decisions is which dimensions to request. Returning 130 dimensions when you only need 4 wastes bandwidth, parsing time, and potentially request quota. But requesting too few means missing data you’ll need later.

Most body measurement APIs organize their output into anatomical bundles — logical groupings of related dimensions that correspond to body regions. Understanding what’s in each bundle helps you choose the right one for your use case without over-fetching.

The four anatomical bundles

FULL_BODY — 130 dimensions

The complete output. Every dimension the model can predict, across all anatomical regions. Use this when:

• You’re building a general-purpose body profile store
• You don’t know in advance which dimensions downstream features will need
• You’re debugging or exploring what the API returns

"requested_dimensions": { "bundle": "FULL_BODY" }

This is the default if no bundle is specified. For production integrations with clear use cases, you almost always want something smaller.

TORSO — 29 dimensions

The most commonly requested bundle for e-commerce and fashion applications. Covers:

• Chest circumference (bust, chest girth, chest breadth)
• Waist measurements (natural waist, omphalion level, iliac crest)
• Hip measurements (maximum hip, seat circumference)
• Shoulder dimensions (biacromial breadth, shoulder width, shoulder slope)
• Torso lengths (sitting height, trunk length, shoulder-to-crotch length)
• Neck and collar dimensions

"requested_dimensions": { "bundle": "TORSO" }

Use case: fashion sizing. Chest, waist, hip, and shoulder dimensions map directly to size charts for tops, jackets, dresses, and trousers. The TORSO bundle covers all of them without returning unrelated foot or hand data.

HAND_ARM — 32 dimensions

Covers the upper limb from shoulder to fingertip:

• Hand dimensions (length, breadth, palm width, finger lengths and widths)
• Wrist circumference
• Forearm and upper arm lengths
• Arm span and reach measurements
• Grip diameter estimates
• Elbow breadth

"requested_dimensions": { "bundle": "HAND_ARM" }

Use case: gloves, sleeves, and ergonomic tools. If you’re sizing work gloves, measuring sleeve length, or designing hand tools, HAND_ARM has what you need. For VR controller ergonomics or wearable device bands, the wrist and hand dimensions are the relevant subset.

LEGS_FEET — 34 dimensions

Covers the lower body from waist to toe:

• Leg lengths (inseam, crotch height, hip height, knee height, sitting height)
• Foot measurements (foot length, foot breadth, heel-ball length)
• Ankle and calf circumferences
• Thigh circumference and thigh length
• Gluteal furrow height

"requested_dimensions": { "bundle": "LEGS_FEET" }

Use case: trousers, footwear, sports equipment. Inseam length and thigh circumference for trousers. Foot length and breadth for footwear sizing. Ski boot fitting typically uses calf circumference, ankle circumference, and foot length together.

HEAD_FACE — 20 dimensions

The most specialized bundle:

• Head circumference and head dimensions (sagittal arc, bitragion arc, biparietal breadth)
• Face dimensions (face width, chin height, face length)
• Interpupillary distance (IPD)
• Ear dimensions (ear length, ear protrusion)
• Neck circumference

"requested_dimensions": { "bundle": "HEAD_FACE" }

Use case: helmets, eyewear, VR headsets, hearing aids. Head circumference for hat and helmet sizing. Biparietal breadth for safety helmet selection. IPD is essential for VR headset optical calibration — it’s one of the most impactful dimensions for VR image quality. For the pediatric model, HEAD_FACE is the primary output bundle because head circumference is the most accurately predicted pediatric measurement.

Requesting specific dimensions

Bundles are all-or-nothing. If you only need inseam length and chest circumference — dimensions from two different bundles — you don’t need to request both full bundles. Use specific_dimensions instead:

"requested_dimensions": {
    "bundle": None,
    "specific_dimensions": [
        "chest_circumference",
        "waist_circumference_natural",
        "hip_circumference",
        "inseam_length"
    ]
}

This returns exactly those four dimensions. The response payload is minimal, parsing is fast, and you’re not requesting data you don’t need.

The specific_dimensions list accepts any dimension key that exists in the model output — you can mix dimensions across bundles freely.

Bundle selection by use case

Confidence scores vary by bundle

One thing worth knowing: requesting a narrower bundle doesn’t improve confidence scores. Confidence is determined by what you input (anchor tier), not what you request. Providing height, weight, and a hand circumference measurement will improve HAND_ARM dimension confidence regardless of whether you’re requesting the full bundle or just hand dimensions.

The exception is when you provide a bundle-specific anchor. If you’re only requesting HEAD_FACE dimensions and you provide head circumference as an input anchor, that anchor will improve confidence for HEAD_FACE dimensions specifically, while having no effect on TORSO or LEGS_FEET confidence.

The pediatric model and bundles

For pediatric predictions (age ≤ 20, or any of the pediatric age categories: INFANT, TODDLER, CHILD, PRE_TEEN, TEEN), the bundle selection follows the same API, but the available dimensions differ.

LMS-derived dimensions (height, weight, head circumference) carry confidence 99 for pediatric subjects — they’re directly calibrated from CDC/WHO growth charts. Ridge-scaled dimensions (most other measurements) are capped at confidence 80.

For pediatric use cases, HEAD_FACE is the highest-quality bundle. Full-body pediatric predictions are available but carry lower confidence outside the LMS-calibrated dimensions.

Choosing the right bundle at the architecture stage saves you from unnecessary data processing later. The rule of thumb: start specific, expand to a full bundle only when your use case genuinely requires it.