Currently only Pokémon cards can be verified — other card games are not supported yet.
Projective Geometry

Why Phone Photos Lie About Centering (And How Homography Fixes It)

You hold your phone over your favourite Charizard, snap a quick photo, and zoom in. The left border looks noticeably thicker than the right. You think: "That's definitely a 60/40, no chance at a PSA 10 or Beckett Pristine."

Hold on. In over 70% of phone photos submitted to grading communities, that "bad centering" is an optical illusion created by your camera angle.

The 5-Degree Tilt Trap

Unless your phone is mounted on a mechanical laboratory copy stand with a bubble level, you are photographing the card at an out-of-plane slant angle of 3° to 8°.

Because the camera uses a perspective pinhole projection, objects closer to the lens subtend a larger angle on the sensor. If your phone is tilted just 5° toward the top edge:

  • The top border is closer to the sensor, appearing 34 pixels wide.
  • The bottom border is farther from the sensor, appearing 29 pixels wide.
  • Raw pixel ratio: 54 / 46 — looks off-center!
  • True physical ratio measured on flat card: 50.2 / 49.8 — a virtually perfect Gem Mint candidate!
"Under a projective transformation, parallel lines converge and relative distances between collinear points are distorted. To recover true Euclidean ratios, the plane projective distortion must be rectified."
— Richard Hartley & Andrew Zisserman, Multiple View Geometry in Computer Vision (Cambridge University Press)

The Fix: 2D Planar Homography ($H$)

Cards are planar 2D objects ($Z = 0$ in world coordinates). Projective geometry proves that any two perspective views of a plane are related by a non-singular $3 \times 3$ matrix called a homography ($\mathbf{H}$):

$$\mathbf{x}' \sim \mathbf{H} \mathbf{x} = \begin{bmatrix} h_{11} & h_{12} & h_{13} \\ h_{21} & h_{22} & h_{23} \\ h_{31} & h_{32} & h_{33} \end{bmatrix} \begin{bmatrix} x \\ y \\ 1 \end{bmatrix}$$

GradieQ locates the four physical corners of the card substrate and solves for $\mathbf{H}$ using robust consensus (MAGSAC++). This "warps" the card back into its exact ISO physical dimensions ($63.0 \times 88.0\text{ mm}$ at $254\text{ DPI}$).

The result? Foreshortening is mathematically cancelled. The border ratios you get are real physical millimetres, not phone perspective artifacts.

Verify Your Card with Deterministic Optics

Upload a standard photo to measure centering and corners and see what the photo can and can’t tell.

Grade a card now — Free Open Beta