Skip to content

UI Based Multi-Person Face Image Search with Ranked Results #1302

@rohan-pandeyy

Description

@rohan-pandeyy

Part of the GSoC 2026 proposal: Notion link

Summary

PictoPy can already cluster face identities using DBSCAN, but the current search interface only supports querying for a single person at a time. This issue tracks the implementation of multi-person face search, allowing users to select multiple identities and retrieve photos ranked by how many of the selected people appear together.

Motivation

The single-person search workflow breaks down when a user wants to find photos of a specific group — for example, finding family photos where both parents and a child are present. There is no way to express that query today without manually intersecting results from separate searches.

What's being implemented

UI

A Search by Multiple People button is added to the top-right of the Face Collections card on the AI Tagging page. Clicking it opens a modal where:

  • All known face cluster identities are displayed as selectable avatar cards
  • The user selects one or more people
  • A match mode toggle controls whether results must contain any selected person or all of them
  • Selected identities appear as dismissible chips above the grid
  • Clicking Search queries the backend and updates the gallery with ranked results

Photos are ordered by match_count descending — images where more of the selected people appear together surface first.

Screenshot

The final implementation would look something like this:
Image

Metadata

Metadata

Assignees

No fields configured for Feature.

Projects

Status
In Progress

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions