What the camera is actually seeing
A photograph is a 2D projection of a 3D body. Software can infer depth from shading, limb overlap and known proportions, but it is reconstructing something the camera never captured. That is the honest framing: photogrammetry is inference, not measurement.
The underlying model
Most of these systems run on a statistical body model. The dominant one is SMPL, published by researchers at the Max Planck Institute and Industrial Light & Magic in 2015. It represents a body as 6,890 vertices across 13,776 faces, driven by 23 joints.
Shape is controlled by a small set of coefficients — ten by default — derived by running principal component analysis across a database of body scans. In plain terms, the model assumes most real human shapes are variations along about ten independent dimensions. The first eight components capture 96.6% of the variance across 2,000 bodies.
That is why photo-based measurement is plausible: human shape really is low-dimensional. It is also why the ceiling exists — unusual proportions get pushed toward the nearest average body.
Where it breaks down
Recent validation work on whole-body smartphone photogrammetry reports encouraging results under controlled conditions but weaker ones on people. Error exceeded 25 mm at the bust, underarm and crotch. Critically, the human portion of that study did not include direct tape comparison at all.
The authors of that work noted that existing validation studies have almost exclusively examined small anatomical regions on relatively small datasets. Put plainly: there is no rigorous published benchmark for photo-based AI estimation of chest, shoulder, neck, inseam, thigh, calf and back length — the measurements that actually determine garment fit.
If a provider quotes you a precise figure for inseam accuracy from photos, ask for the study. It probably does not exist.
Photo conditions that matter more than you would think
- Wear something close-fitting. Loose fabric destroys silhouette inference.
- Keep limbs away from your torso. Arm-to-body contact hides the waist boundary.
- Stand naturally with weight even. A contrapposto stance skews everything.
- Shoot in even, diffuse light. Hard shadows read as body shape.
- Use a plain background and remove loose hair or accessories.
The verdict
Photo-based measurement is genuinely useful and getting better fast. It is not a tape measure, and anyone presenting it as one is overselling. For broad proportions — height, shoulder width, rough build — it is good. For the tight tolerances that a fitted garment demands, treat it as a starting point and verify the critical measurements yourself.