Batch Label Recognition for Samples
Feature in One Sentence
Lay out multiple fabric swatch cards in a single photo, let AI recognize them all at once, then create samples in bulk after reviewing the results. It eliminates the repetitive work of photographing and entering data one card at a time, making new samples enter your library several times faster.
Difference from "Single-Image AI Recognition"
| Scenario | What to use |
|---|---|
| Photographing a single swatch card / label | AI Fabric Label Recognition (single-image mode) |
| Multiple swatch cards in one photo / a tableful of cards shot at once | This feature: Batch Label Recognition for Samples |
| Legacy data in an Excel sheet | Batch Import (no AI involved) |
How to Use
Entry Point
PC → left menu "Sample Library" → at the top-right of the sample list page, the "Batch Label Recognition" button next to "Batch Import" (camera icon).
Steps
- Click "Batch Label Recognition" → upload images
- Upload one or more photos
- Each photo can contain multiple fabric swatch cards
- The system automatically crops and recognizes each card one by one
- Choose the source: are these samples "your own products" or "provided by a supplier"?
- When you choose "Supplier", the recognized "code" is kept as the supplier code
- Wait for the recognition results (usually 5–10 seconds per image)
- Preview + deduplicate
- Each recognized swatch card = one row of a pending sample
- Fields: code / name / composition / width / weight / color, etc.
- Duplicate codes are highlighted in red; the system automatically keeps the first row of each group, and you can manually uncheck the extra rows
- Create samples → automatically jumps to the "Batch Edit" page, where you review each row, fix fields, remove duplicates, and confirm before saving to your library
Photography Tips (the key to accurate recognition)
Take this photo well for a high recognition rate
- Lay the cards flat—don't stack or overlap them
- Plenty of light—don't shoot in a dim spot
- Keep the cards face up so the text is clearly readable
- No more than 6 cards per photo is recommended; too many will crowd each other
- Don't use too low a resolution; normal phone clarity is fine
A Few Details About Reviewing
Duplicate codes are highlighted in red
If duplicate codes are recognized from the same batch of photos, the preview table highlights the duplicate rows in red as a reminder (duplicates are easy to recognize when cards are stacked or obscured). The system automatically keeps the first row of each group and unchecks the extra rows. If different cards were genuinely recognized as duplicates, go back and re-shoot them.
Missing fields are left blank
When AI doesn't recognize certain fields, they are left blank. You can:
- Fill them in manually on the "Batch Edit" page after creating the samples
- Or create the samples first and edit them later
Unchecking / cannot add rows
- Unchecking: a row was recognized incorrectly / is a duplicate → uncheck it so that sample isn't created
- Cannot add rows: for samples not recognized this time, use "Single-Image AI Recognition" or enter them manually after saving
Your Own Products vs. Supplier
When doing batch recognition, choose one source for the entire batch:
| Source | Where the code goes |
|---|---|
| Your own products | The AI-recognized code → written into the "own code" field |
| Provided by a supplier | The AI-recognized code → written into the "supplier code" field; the own code is left blank for you to fill in |
You can only choose one source at a time. If a batch of photos contains both your own products and supplier cards, recognize them in two separate runs.
FAQ
Q: How many fields can be recognized? A: The core fields—composition, width, weight, color, code, and name. Exactly which ones are recognized depends on what's printed on the card; AI recognizes whatever it sees.
Q: How accurate is recognition? A: Cleanly printed cards have a high recognition rate; handwritten, blurry, obscured, or glare-affected cards have a lower rate. That's why the review table must always be checked by a human before saving.
Q: How is the cost per recognition calculated? A: It's billed by AI usage. See AI Top-Up and Cost-Saving Guide.
Q: Can I use it in the app? A: This feature is launched on the PC first. Single-image "AI Fabric Label Recognition" is still available in the app—for field staff photographing a single card, single-image recognition is more convenient.
Q: Recognition failed / the result is way off? A: First check whether the photo is unclear (lighting / obstruction / blur). Re-shooting usually gets accurate results. For systematic recognition bias, please contact support to report it—we keep improving the model.
Related Articles
- Sample Management — includes single-image AI fabric label recognition
- Batch Data Import — Excel batch import (no AI involved)
- AI Top-Up and Cost-Saving Guide — AI usage billing
- Image Search — photograph a fabric to find similar samples in your library