Pole Site CAD
Parametric model of the three-pole site, fitted from the 20 phone photos. Tapered poles, crossarms, risers, insulators and conductors; the geometry itself uses no photos or point cloud. Optional layers overlay VGGT’s points and the photos so you can check the fit. Drag to orbit, pinch to zoom, tap a part for its dimensions.
How reliable is this
- Camera positions were re-solved from matched photo features (about 1 px error) because VGGT's own camera estimates were several degrees off. Poles were then fitted to the photo outlines and agree with them to about 3 to 5 px.
- Pole positions, diameters and crossarm heights come from that fit and are the strongest part. Poles are drawn plumb with one shared taper. Pole tops and ground levels are estimates, so pole heights can be off by a few percent.
- Crossarms sit at two measured azimuths (along the straight span and along the turn) with one shared length, centred on the pole.
- Firm conductors are the three wires between poles 1 and 2, triangulated from many photos and attached at pin or hanging insulators. Dashed ones have weaker evidence. The span between poles 0 and 1 is only partly resolved because the photos look along it. Wires are straight chords with no sag.
- Pole 0 and pole 1 risers were triangulated from several photos; the pole 2 riser was seen in only one.
- VGGT points are VGGT’s raw depth, one photo at a time, moved onto that photo’s corrected camera. VGGT’s depths ran 7 to 13 percent long against the fitted poles, so each photo’s depths are rescaled once to match them. Even so the points are noisy at pole scale (about half a pole radius on the poles, worse at edges and in trees), so use them for context and for checking, not for measuring. Look-through mode shows each photo with the model projected over it using the same corrected camera.
- Nothing here is survey-grade. Units are relative until you set a scale from a known length, such as a crossarm or pole height.