hmgill's picture
Upload 71 files
876cc3e verified
|
Raw
History Blame Contribute Delete
1.2 kB

Example specimens

One folder per specimen. Each folder is 1–4 photographs of the same fossil from different angles — that is the pipeline's unit of work, and Meshy's hard limit.

examples.csv columns:

column
id folder-safe identifier
title shown on the gallery card
folder path to the view folder, relative to the Space root
prompt prefilled into the composer when the card is clicked (editable)
source attribution

Honest note on the bundled example

IU-3105/ is four rotations of one photograph, not four real angles. It exercises the plumbing — the gallery, the multi-view folder path, the per-view segmentation loop — and it will produce a mesh. That mesh will not be a good reconstruction, because rotating one image adds no new information about the specimen: Meshy still cannot see the back.

Replace it with genuine angles before drawing any conclusion about output quality. Write the prompt to say something specific about that specimen — the shadow, the broken face, the matrix still attached — because the prompt is what steers the analyst's keypoints, and a generic one wastes the best lever you have.