After toiling for a few months on this, product image classification is now live on Datagene.io! While the product classification API works with product titles, the image classification API works with product images, though only for fashion.

Update: API discontinued to save on cloud cost.

This is part of a series of posts on building a product classification API:

• Data acquisition and formatting (part 1)
• Data cleaning and preparation (part 2)
• API development (part 3)
• Image classification demo
• Image search demo

Some facts about the image classification API:

• Works best with e-commerce like fashion images (as that’s what it was trained on)
• Top-1 validation accuracy: 0.76; Top-5 validation accuracy: 0.974
• Returns results under 300 milliseconds (will be faster in batch mode with GPU)
• Built on Keras and Theano, and runs on a tiny AWS server without GPU.

How to play with it? First, click on browse to select a file for upload. The file needs to have either a “.png”, “.jpg”, or “.jpeg” extension (case-insensitive).

A very simple input form for your images, with basic validation.

Then, click on the submit button below and wait patiently—results should appear in a second or so, depending on network speed and image size.

A very simple submit button.

Here’s an example

Given a photo of a t-shirt…

I wonder if the model will also have learnt about the tagline on the t-shirt.

… here’s the result we get.

Oh, they don't have a category for smartass t-shirt.

It works on mobile too

Look ma, web responsive!

You can also try the image classification API on mobile. On iOS, it allows you to take a photo, or select one from your photo library or iCloud drive.

Please leave suggestions on UI (or any other) improvements in the comments!

Related writing

• Product title classifier: Data acquisition and formatting
• Product title classifier: Data cleaning and preparation
• Product title classifier: App development
• Product image search demo