Automatic camera path, leaning towards the photographs
The automatic path still chooses its stations in free space by photo support, interest and openness. Photo support now weighs more (power 1.5 instead of 1) and a station may come within 0.2 m of a photograph instead of 0.35 m, so the flight keeps closer to the views the photographs took without stepping from photograph to photograph.
One scene, both paths
Same reconstruction of the rotunda museum, same seed, two chunks. On median the camera is 0.75 m from the nearest photograph on the previous path and 0.63 m on the new one.
Previous path: proxy (left) and generated video (right)
Context photographs of the first chunk
New path: proxy (left) and generated video (right)
Context photographs of the first chunk
Thirteen scenes, paths only
Three chunks per path, seed 0, on the thirteen scene bundles used to develop the automatic path; no video generated.
| previous | new | |
|---|---|---|
| Median distance from the camera to the nearest photograph | 1.05 m | 0.94 m |
| Frames close to a photograph in position and direction | 45% | 52% |
| Coverage of the views by the context photographs | 0.708 | 0.726 |
| Paths free of collisions | 13 of 13 | 13 of 13 |