AIIMAGESUITES

USE CASE · AI / ML & COMPUTER-VISION TEAMS

COLLECT LABELLED IMAGES,
NOT A LABELLING BACKLOG.

Capture training images with the class label, split and session written into the filename and EXIF at the moment of capture. Sequenced names mean deterministic ordering; tags in XPKeywords and UserComment mean the label survives every copy, resize and re-export until it reaches your loader.

Start Capturing

WHAT GOES WRONG TODAY

  • Label after the fact → labelling backlog grows faster than the dataset
  • Filenames get normalised by tooling; sidecar CSVs get lost
  • Class imbalance is invisible until you count files by hand
  • Multiple collectors, multiple phones, zero naming consistency

WHAT GETS BAKED INTO EVERY FILE

  • ImageDescription — session name + tags
  • XPKeywords — tags (Windows / Lightroom keyword search)
  • UserComment — full capture record
  • Artist — session name
  • DateTimeOriginal / DateTimeDigitized — capture time
  • Software — AIImageSuite.pro / Session Capture

THE WORKFLOW

  1. 01

    One session per class

    Name = class label; tags = split, environment, collector.

  2. 02

    Capture in bursts

    CLASS-0001.jpg… tags baked into EXIF. Counter shows how many you have per class.

  3. 03

    Switch class

    Duplicate the session, change the label, keep going — balance as you shoot.

  4. 04

    Export

    ZIP per session, or Cloud Sync. Read labels back with exiftool / piexif in your pipeline.

QUESTIONS PEOPLE ASK

How do I read the labels back?

exiftool -ImageDescription -XPKeywords -UserComment *.jpg, or piexif / Pillow in Python. The same text is also in the filename.

Is capture resolution controllable?

Captures use the camera's native stream resolution offered by the browser; the camera diagnostic on the setup page shows what each device delivers before you start.

Does it support multiple collectors?

Yes — each collector signs in and captures to their own sessions; the naming scheme is enforced by the session so files stay consistent across phones.

Can I capture offline in the field?

Yes. The PWA captures offline and resumes uploads when connectivity returns.