no trimming of pointcloud/mesh no use of markers, all 64 pics aligned correctly in photoscan decimated to 1m from 3.4m polys \1x 8k texture notes: the dataset only contained 64 pics or 64 cameras, I was expecting 72 several pics are repeats>virtually the same shots several pics are NOT in adequate focus jpgs do not have the same quality for matching points as RAW all photos aligned first go with photoscan, not with reality capture quite alot of cleanup required to make this model complete and printable :( alot of noise from lack of sharp details in focus and missing points (probably due to the lack of cameras Verdict: the supplied dataset is sub-par quality, with duplicates, unfocused shots, sub optimal array configuration, and lacks the number of nodes the rig boasts (64-duplicates vs 72) so therefor is not an measurable dataset to inspecting and rating functionally. JPGs can only replace RAW for high quality reconstruction if the shots are sharp and in focus, and or if bandwidth does not permit
Comments