.png)
Photo Batch Analysis
Discover and cluster all unique individuals across an entire image folder
Photo Batch Analysis (internally called Folder Identity Lab) processes a folder of images and produces a person-first view of the entire collection. Instead of searching for one specific person, it finds everyone — clustering all faces it detects into unique identity groups and showing every photo in which each person appears.
Use Cases
- Event photography — identify all distinct individuals across hundreds of event photos.
- CCTV archive review — find out who appeared across a day's worth of surveillance snapshots.
- HR and access records — audit a dataset of employee photos to verify identity coverage.
How It Works
- Select an image source — either browse a local folder from the browser, or select a pre-indexed server-side folder.
- Start Analysis — images are processed as a background job. The system detects all faces in every image and computes ArcFace embeddings.
- Clustering — faces are grouped by identity using embedding similarity. Known faces (those in the enrollment database) are labelled by name; unknown faces are grouped into anonymous clusters.
- Person-first results — the output is organised by person. Click any person's tab to see every photo in the collection where they appear.
Image Sources
| Source | How to Use |
|---|---|
| Local Folder | Click Browse Folder and select a directory from your device |
| Server Folder | Select a pre-indexed directory from the dropdown (set up by the backend administrator) |
Result Summary
After analysis completes, the summary panel shows:
| Stat | Description |
|---|---|
| Unique People | Total number of distinct identity clusters discovered |
| Total Faces | Sum of all face detections across all images |
| Known Matches | Faces matched against enrolled identities |
| Processed Images | Total number of images scanned |
| Time Taken | End-to-end processing duration |
Best Practices
- For best clustering accuracy, ensure images are reasonably sharp and faces are at least 64×64 pixels in size.
- Adding more enrolled identities improves the ratio of named vs. anonymous clusters.
- Large folders (1000+ images) benefit from GPU acceleration on the backend — processing time scales linearly with image count.
- Use the Reset button to clear state before starting a new analysis session.