AI Tagging
Every item you add goes through an AI vision analysis that generates a structured tag set. This is how the suggestion engine understands your wardrobe well enough to build outfits.
How the Pipeline Works
- Upload: You submit a photo, URL, or text description.
- Processing badge: The item appears in your wardrobe grid immediately with a “Processing” label.
- AI vision analysis: The backend sends the image to an AI vision model. It identifies clothing type, color, style signals, material indicators, and season fit.
- Tags generated: The model outputs a structured set of tags (see table below).
- Background removal: The background is stripped from the photo to produce a clean thumbnail.
- Complete: The processing badge clears. The item now has its full tag set.
What Gets Tagged
| Tag | Example values |
|---|---|
| Type | shirt, trousers, dress, jacket, shoes, bag |
| Subtype | t-shirt, chinos, midi dress, bomber jacket, chelsea boots |
| Color | navy, white, olive, burgundy, black |
| Secondary colors | e.g., white stripes on a navy shirt |
| Style | casual, formal, sporty, minimalist, bold |
| Formality | casual / smart casual / business casual / business formal / black tie |
| Season | spring, summer, autumn, winter, all-season |
| Material | cotton, wool, denim, leather, polyester, linen |
| Pattern | solid, striped, checked, floral, geometric, animal print |
Accuracy Notes
AI tagging is accurate for clearly photographed items. It performs best on:
- Solid colors and simple patterns
- Standard clothing silhouettes
- Well-lit photos with minimal clutter
It can struggle with:
- Complex patterns on dark backgrounds
- Items heavily obscured by folds or wrinkles
- Accessories with fine detail (jewelry, ties)
- Items photographed as part of a complete outfit (as opposed to isolated)
Confidence scores below 70% on any tag are worth reviewing.
Re-Analyzing Items
If an item was tagged poorly (wrong color, wrong type), you can trigger a reanalysis from the item detail view using the overflow menu. This re-runs the full pipeline including background removal. Existing manual edits to tags are overwritten by the new analysis, so re-check after reanalyzing.
Tips for Better Photos
- Lay items flat on a light, neutral surface or hang them on a plain hanger.
- Photograph the front of the item straight on, not at an angle.
- Use natural light or diffused indoor light. Avoid harsh shadows.
- Keep the entire item in frame with a small border of background visible.
- For shoes, photograph one shoe from the side.
- Avoid photographing items while being worn. The system analyzes the garment, not the body.
Custom AI Endpoint
If you’re self-hosted or want to use a specific model, you can configure a custom AI endpoint in AI Configuration. The tagging pipeline will route through your endpoint instead of the default.
Mobile: AI tagging works identically on mobile. After uploading from the camera or photo library, the processing badge appears in the wardrobe tab until analysis completes.