How Many Images per Sequence?

Imagery, data, and integrations Imagery
11

@paulinus @balys I have a few more questions about the correlation between images per sequence, space density, and Mapillary reconstruction accuracy.

  • Do you favor image resolution or camera position density (number of image per unit of volume), or both? @boris has suggested that Mapillary targets ⅓ images per meter with videos, which seems to me rather a bit sparse for high density maps. Does then going denser than ⅓ images per meter mean a waste of effort? Bee Maps goes for 30 fps in their capturing devices (for cars), which equates to denser coverage per meter when capturing by car at urban speeds. I am though aware that Bee Maps may not be using every frame for reconstruction either.
  • Is there a minimum space density resolution (camera positions per unit of volume) in Mapillary’s reconstruction?
  • Some of my sequences fail due to not reconstructing and I do not know why:
    9bO4gcxKUn7GHIDRAQ0Mtm
a6OP2L8gFd0TXIYVQyjEhD
wo8HOv2YmbLVuRl4WngDe7
GbzQ908mheDJPL76cIXstF
btm3hiTFOWzqdyunCK7aBG
1 Like