This model is best viewed by selecting the first person mode in navigation setting and slow speed (eg 1 or 2) and navigating at the neighborhood of the small cones. It is computed from 343 1632x1224 images taken during a complete walk around a church. The cones indicate camera locations and view field of the camera.
The camera is catadioptric (the 0-360 mirror is mounted on Nikon Coolpix 8700), it is hand-held and mounted on a monopod. It has a large view field: 360° in the horizontal plane and about 50-60° above and below.
Both calibration and structure-from-motion are computed as in the CVIU‘08 paper. The surface is estimated by an extended version of the method in the CVIU‘13 paper: curve points are (matched by the method in the PAMI‘02 paper and) also reconstructed, not only interest points.
More details and videos for similar models at a larger scale are available from my web page.
Comments