Similar-photo groups, selections, and cache

Compare the similar. Keep the distinct.

Keep the top 1–9 qualifying photos per conservative group. Each group retains at least one; if none passes the threshold, the highest-ranked photo is kept for manual review. Capture time, camera information, and visual geometry help separate groups.

Use a 30-second, 2-minute, 10-minute, or unlimited capture interval. After analysis, merge groups, split photos into separate groups, or edit group numbers inside Lightroom. Saved revisions reuse existing scores without rerunning AI.

Groups you can adjust

After a Best per View analysis, open Results / Groups / Cleanup → Adjust last groups. Matching group numbers mean the same group. Select photos across pages to merge them, or make each one independent. Save a separate revision, then review it through View last results. This reuses existing scores instead of exporting or scoring again.

Reuse the analysis you already have

Changing thresholds, retention counts, or grouping settings can reuse scores and features. A complete cache hit skips JPEG export and AI scoring. Photo or develop changes invalidate the cache; the numeric and feature cache is capped at 256 MB.

Can it identify front, side, and rear views?

It cannot reliably identify those semantic view categories. The plugin uses feature matching and geometric checks to form conservative groups. Uncertain matches stay in separate groups, and photos in different groups do not compete.

If photos or develop settings changed after analysis, run a new analysis before editing groups. Older results without the required grouping data also need a fresh run. A low-quality fallback keeps a candidate for review; it does not certify image quality.

What has been validated so far?

V1.3 has recorded native runtime, grouping, cache, cleanup, and multilingual regression tests. Settings windows have been checked in Lightroom. Full grouping edits were mainly tested with SDK simulations and an eight-photo pipeline; real catalog ratings were not changed in these tests. Start with a small set.