unthinkable
Case study · eJohri, India · omnichannel jewellery marketplace

Jewellers join eJohri and start selling sooner, because their product photos are cleaned in seconds.

eJohri wanted jewellers from cities and towns across India to get their products in front of buyers quickly, looking their best.

IndiaBrand storefronts

What changed

2-3 secondsto edit an image, instead of hours by hand (as reported)
98%increase in image editing efficiency (as reported)
95%model accuracy (as reported)
100,000+images analysed and corrected (as reported)
What stood in the way

What stood in the way.

What stood in the way
  • Every jeweller's product photos were cleaned by hand, removing backgrounds, shadows and stray objects, which took hours per image.
  • That slowed onboarding as the marketplace grew, and the training data available was limited.
What changed
  • 2-3 secondsto edit an image, instead of hours by hand (as reported)
  • 98%increase in image editing efficiency (as reported)
  • 95%model accuracy (as reported)
  • 100,000+images analysed and corrected (as reported)
What we built

What we built.

  • Tests of computer vision techniques (edge detection, thresholding, K-clustering), dropped because results varied by type of jewellery.
  • A machine learning model trained by transfer learning on a general jewellery data set, to make up for the small data set.
  • Bulk editing that keeps the image quality exactly as the jeweller supplied it.
  • Machine learning (transfer learning)
  • Computer vision
What it shows

What this work shows about us.

We pick the method that works, then build it

We tried simpler techniques first and moved to a trained model only when they failed on real jewellery.

Onboarding is a sales problem

Seconds per photo instead of hours means a jeweller is selling sooner.

Want a similar machine learning model for editing product photos? Tell us what you want to change. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.

Start your plan ↗
If we built it today

If we built it today, with AI.

What we would add now, moment by moment.

Finding it

An agent would write each listing's title, metal, purity and description from the jeweller's photos and price sheet, for the jeweller to approve before it goes live.

Choosing in store

An agent would answer a buyer's question about a piece and book a visit to the nearby jeweller who stocks it. The jeweller still confirms the visit and the price.

Ask Unthinkable

Want this for your business?

Tell us what you run or what you're building. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with a senior AI engineer.

Send us a note instead

Required

Thanks. A senior AI engineer will reply to you.
u
Ask UnthinkableClarity first, then the right team
Your conversation

An AI agent. This prototype uses prepared answers and sends nothing.