Duplicate Finder Keep Rules & Weights

Finding duplicates is half the job

The Perceptual Duplicate Finder groups visually similar images for you. The more interesting question is what happens next: out of each group, which copy do you keep, and what do you do with the rest? Bulk Image Downloader From URL List answers that with keep rules, weight tuning, group actions, and the ability to export the duplicate URLs. This is where a scan turns into a cleaned list.

Keep rules: choosing the survivor

A keep rule tells the finder which image in each group is the one to keep. You set it before or alongside the scan, and the options cover the decisions people actually make:

Largest file — keep the highest-quality copy.
First in list — keep the earliest occurrence.
Smallest — keep the leanest file when size matters more than fidelity.
Shortest URL — often the cleanest, most canonical source.
Manual — decide group by group yourself.

With a manual rule, you pick the keeper from a dropdown on each group, or use Keep This on any image to promote it directly. That flexibility matters when an automated rule gets it right most of the time but you want the final say on a handful of groups.

Tuning the weights

The finder combines many signals to judge similarity, and you can adjust how much each one counts. Expand the weight tuning controls, change the signal weights, and use Re-apply to re-evaluate the existing results — no full rescan required. That fast feedback loop is the point: nudge the weights, re-apply, see whether the groups tighten up or loosen, and repeat. If you push it too far, Reset to defaults puts you back to a known starting point so you can tune again until the groups feel right.

Acting on groups

Once the groups and keepers look correct, the action tools let you move fast. Per group you have Select Group, Select Non-Kept, Download Group, and Remove Non-Kept. Per image, you can Copy URL or Open from the row. An action bar adds selection helpers across the whole report — select all, keepers only, non-kept only, or clear the selection.

From there you act: remove the selected URLs from your tasks, remove the non-kept copies according to your keep rule, or download the selected or non-kept images for review before deleting anything.

Exporting the duplicates

You are not limited to removing duplicates in place. The finder lets you Export Duplicate URLs, copy them, or open them in tabs. That is useful when you want a record of what was flagged, need to hand the list to another tool, or want to double-check the duplicates externally before pruning. Between keep rules that automate the obvious choices, weight tuning that sharpens the matching, and group actions that clean up in bulk, the finder takes you from a raw scan to a deduplicated, download-ready list with very little manual sorting.