From field photographs to annotation-ready settlement-plate imagery — in the browser.

Versão em português → README.pt.md

PlatePrep prepares settlement-plate campaign photographs for annotation in CoralNet and other platforms, in a single guided browser workflow: campaign metadata → per-plate triage → selection → perspective-corrected cropping at fixed physical scale → named images + ready-to-import CSVs + a crop-geometry manifest.

• Single file, no installation — PlatePrep.html runs in any Chromium-based desktop browser.
• Bilingual interface (English / Portuguese), switchable at any time.
• Photos never leave your computer — all image processing is local. The only optional network call is the weather lookup (Open-Meteo).
• Reproducible by design — every crop's four source corners are recorded in a manifest, so any crop can be re-derived and operator error can be screened without re-inspecting images.

Try it now: https://www.projetobioe.com/plateprep.html (hosted instance, HTTPS)

The 5-step workflow

Step	What happens
1 · Set up	Experiment, site, photographer, depth, number of plates, plate size and the plate → treatment map. Date, camera and GPS are read from the photos' EXIF; without GPS (e.g. GoPro HERO12), type coordinates or pick a saved site. One click fetches the day's weather from Open-Meteo (air temperature and conditions from its historical service, based on the ERA5 reanalysis; water temperature from its marine service). Everything is editable.
2 · Triage (optional)	The campaign photos in sequence with EXIF time; click the first photo of each plate and the program groups and numbers the rest. Can be skipped.
3 · Selection	Thumbnail gallery; each photo shows its triage plate; check the ones to crop (already-cropped photos start unchecked).
4 · Crop	Click the plate's 4 corners → exact per-pixel homography rectification at fixed scale (image side = plate side), from the photograph at native resolution. The plate is pre-filled from triage; treatment, file name and sequence are automatic. Enter saves and advances.
5 · Finish	The output folder receives the crops (Plate##_T?_EXP_YYYY_MM_DD_##.JPG, scale embedded in the JFIF header), metadata_coralnet.csv (import-ready), triage_mapping.csv, crop_manifest.csv/.json (the 4 corners of every crop) and PlatePrep_ImageJ_scale.ijm (ImageJ calibration macro).

The workflow, in pictures

1 · Set up — campaign metadata pre-filled from EXIF/GPS; one-click weather (Open-Meteo).	2 · Triage — click the first photo of each plate; the rest is grouped and numbered.
4 · Crop — four corners → fixed-scale homography; plate inherited from triage; name and notes automatic.	5 · Finish — crops, metadata CSV, triage map and crop manifest written to the output folder.

Full instructions, with screenshots, in the manuals: docs/Manual_PlatePrep_EN.pdf · docs/Manual_PlatePrep_PT.pdf.

Getting started

Use the hosted instance (recommended): open https://www.projetobioe.com/plateprep.html in Google Chrome or Microsoft Edge, click Choose folders and start, and pick the input folder (campaign photos, JPG/PNG) and an output folder.

Or run it yourself: download PlatePrep.html and open it in Chrome. Direct folder access uses the File System Access API, which needs a secure context (HTTPS or localhost); opened from file:// the browser may restrict folder access, in which case PlatePrep falls back to packaging its results into a downloadable .zip. To self-host, serve the single file statically over HTTPS — no build step, no backend.

Requirements

• Chromium-based desktop browser (Chrome, Edge). Other browsers fall back to .zip output.
• Photos available on the local disk (in Google Drive, mark the folder Available offline).
• Internet only for the optional weather lookup (and for the .zip fallback, which loads JSZip from a CDN); all fields can be filled in manually.

Outputs

File	Contents
Plate##_T<letter>_<EXP>_<YYYY>_<MM>_<DD>_<seq>.JPG	Square, perspective-corrected, fixed-scale crops (Placa/Plate prefix configurable). Self-describing, machine-parseable names. The physical scale (px/cm) is embedded in the JPEG's JFIF header — nothing is drawn on the image; Photoshop, GIMP, QGIS, Python/PIL etc. open the crop already calibrated.
metadata_coralnet.csv	One record per image with the columns CoralNet imports: Name, Date, Experiment, Site, Treatment, Exposure_Days, Plate, Height (cm), Latitude, Longitude, Depth, Camera, Photographer, Water quality, Strobes, Framing gear used, White balance card, Comments. Exposure_Days is computed from the deployment date; Comments carries the day's weather.
triage_mapping.csv	Original photo → plate / treatment / sequence — a permanent record of the campaign photo log.
crop_manifest.csv / .json	For every crop: the four source-image corner coordinates, working-canvas and source-file dimensions, output size, plate size and px/cm. Re-derive any crop; quantify between-operator variability from the manifests alone.
rederive_log.csv / render_audit.csv	Written by the Re-derive crops from a manifest mode (welcome screen): per-crop source-resolution log and mesh-seam audit — the executable form of the manifest's reproducibility promise.
PlatePrep_ImageJ_scale.ijm	ImageJ/Fiji macro that sets the global spatial calibration of the campaign's crops (Set Scale… distance=<px> known=<cm> unit=cm global). ImageJ's built-in JPEG reader ignores JFIF density, so run this once per session (Plugins › Macros › Run…).

The plate → treatment map of the Bioē coating experiment (30 plates, treatments A–F) is built in as a template and fully editable — including the treatment legend — for any other experiment.

Known limitations (v1.1.2)

• Assumes flat, approximately square plates with all four corners visible; corner marking is manual (automatic corner detection is on the roadmap).
• Weather data lag real time by a few days (ERA5); recent campaigns are filled in manually.
• Fixed in v1.1.0 (see CHANGELOG): photographs are now decoded at native resolution (up to v1.0.5 they were capped at 3600 px wide), and rectification is rendered per pixel through the exact homography (up to v1.0.5 a 26 × 26 mesh left faint seams). Crops made with earlier versions carry both traits; the manifest lets you tell which version produced a crop.

Citing

If you use PlatePrep, please cite the archived software version:

Galembeck, E. & Schlosser, C. F. (2026). PlatePrep: browser-based preparation of settlement-plate imagery for annotation platforms (v1.1.2) [Software]. Zenodo. https://doi.org/10.5281/zenodo.21960482

Concept DOI (always resolves to the latest version): https://doi.org/10.5281/zenodo.21959574 · Machine-readable metadata in CITATION.cff (GitHub's Cite this repository button). A methods paper describing and validating the pipeline is under submission; this section will be updated with the reference.

Contributing

Issues and pull requests are welcome. PlatePrep is a single self-contained HTML file; the rectification (solveH, warpTo), EXIF, CSV and manifest logic are plain JavaScript with no dependencies. Please describe your browser and a minimal photo set when reporting a bug.

License and credits

CC BY 4.0 — see LICENSE.

Bioē Project — Institute of Biology, University of Campinas (Unicamp), Department of Biochemistry and Molecular Biology, with the Centre for Marine Biology (CEBIMar/USP). Funding: FAPESP — Technological Innovation Programs / PROASA – Program for the South Atlantic Ocean and Antarctic Sciences, Research Project – Regular, Call for Proposals (2025), 1st Cycle, grant #2025/0787809.

Weather data: Open-Meteo (historical service based on the ERA5 reanalysis; marine service).