Every model starts with a spark.

An iPhone app to photograph, label and check your images, then train your own vision model.

Coming soon to iPhone

Good evening

You know what your app should see.

A ripe mango. A cracked tile. A missing screw. You need a model that finds it. You do not need a desktop labelling tool or a course in machine learning first.

Open your phone. Start where you are.

Capture

Start with the phone in your pocket.

Take photos of what your model should find. Different light, different angles. Or import the ones you already have.

Photos go to a private folder. Never to your camera roll.

Label

Draw a box. Say what it is.

Auto detect finds the objects on your phone. You tap the right label. Each photo you finish lights up, so you can see the work spread.

46 left · about 14 min at your pace. Stop any time. Your place is kept.

2 blurry

Health check

It tells you what to fix.

Emberlens counts the boxes for each label and shows which labels need more. It also spots blurry photos and photos that look the same, before they waste a training run.

Label counts are free. Blurry and look-alike checks are part of Premium.

Private

Your photos stay yours.

No account. No analytics. No ads. Your photos stay on your phone until you export them, and they go only where you choose.

  • We never see your photos.
  • Auto detect runs on your phone.
  • Imported photos lose their location and camera details.
  • Delete a project, and its photos and labels are gone from your phone.

Export

Take it anywhere.

Save your dataset as COCO, YOLO or Pascal VOC. Send it with AirDrop or Files. Or upload it to your own storage, with your own keys.

Nothing leaves until you tap. Upload to your own storage is part of Premium.

Train

Train it on your laptop.

The free Emberlens SDK opens your export. One command does each job. You do not need to know Python.

It runs on your laptop. Your photos stay with you.

Test

Bring it back. Try it.

Add your model to the app, then try it on new photos. See what it gets right, and see what it needs next.

Using your own model in the app is part of Premium.

fruit-stand-classify-model

CaptureLabelHealth checkExportTrainTestYour model

Your model

Then go round again.

These stages are not steps. Capture ten, label five, capture more. Each round gives your model more to learn from.

Coming soon to iPhone