PlatePrep — validation data, analysis code and regression fixture

Companion deposit to the software record of PlatePrep (browser-based preparation of settlement-plate imagery for annotation platforms) and to the methods paper "PlatePrep: browser-based preparation of settlement-plate imagery for annotation platforms" (Galembeck & Schlosser). Contents. PlatePrep_validation_data.zip: data/crop_manifests/ — the four corner-record manifests behind every number in the paper (519 rows, 479 unique crops; two coating experiments, 14 campaigns, 30 plates each) and the CoralNet metadata table; data/rederived_native/ — the analysed dataset's native-coordinate manifest, per-crop re-derivation log and rendering audit (471 crops; gradient-ratio median 0.99: no rendering trace); analysis/ — reproduce_validation.py, which recomputes every value in Table 2 and Figures 3–4 from the manifests alone, with its expected output; regression_fixture/ — three source photographs, their manifest rows, the crops produced by the earlier build and by the current renderer, and test_rectification.py, an independent re-implementation that asserts agreement with both (6/6 cases pass); software/PlatePrep.html (v1.1.2) and bilingual manuals. source_photographs.zip: the 471 campaign photographs retained on disk. rederived_crops.zip: the rectified crops analysed in the paper (475 JPGs with the physical scale embedded in the JFIF header). The corner records were made with an earlier build preserved under the software concept DOI (10.5281/zenodo.21959574); the analysed dataset was re-derived from those records with the version of record (v1.1.2, DOI 10.5281/zenodo.21960482) using the application's re-derivation mode. Bioē Project, Institute of Biology, University of Campinas, with CEBIMar/USP. Funding: FAPESP – Technological Innovation Programs / PROASA, grant #2025/0787809.