Our imaging CDN in conjuction with our CBIR storage can make any e-commerce outstanding

Imaging CDN

All images in this demo are served by our imaging CDN reducing bandwidth and increasing the speed of the app when using limited connections.

Try some filters.

CBIR storage

Visual Search

Click an image to find similar shoes based on shape, texture and colors.

Step 1.

We indexed the images with shoes' data as a bboxtag. While we stored the main image of every product with type detail the rest were stored with type referer. We ended up with images with two different types of bboxtags.

Step 2.

We created a small API written in PHP that called our _vsearch and _search API. _search is needed to perform searches with no reverse image search involved. _vsearch is the one that finds out similar items based in visual similarities. Both present faceted filters to filter the given results. To create that faceting we did some ElasticSearch's aggregations. While _search returns a document scoring, _vsearch returns a percentage of visual scoring. Note in the _vsearh query example how min_score has been setted to 75% to reflect this property. In the demo we just do visual search by shape, color and texture so we disabled the object matching feature.

Example of _vsearch query

Returned results

Step 3.

To display the product details we obtain all images that belongs to a product just filtering by the bboxtag property data.id_modelo. In imagenii bboxtags, metadata and colors are nested documents that have been also stored flattened.

curl -XGET 'http://www.imagenii.com/imagenii_demo/chachic/_search?_source=id,bboxtags&q=bboxtags.data.id_modelo:252857'

Step 4.

To create the visual recommendation engine found in the product details view we did a _vsearch with the product image filtered by gender and product id and setting min_score to 70%. We could improve the results having more data available like product categories. More complex recommendations engines can be built with imagenii, i.e. modifying how documents scoring can affect visual scoring or query expansion, please refer to the docs to learn more.

Step 5.

That's it!

Augmented Reality

imagenii + louvre database + text to speech API = awesome museum tour

Imaging CDN

The app calls in real-time the version of the image that needs not having to create offline multiple versions that can change in future changes.

Try some filters.

CBIR storage

Just only storing the paintings of a museum we can create not only an awesome experience for the Louvre visitor but also a powerful full text search database that could be used to explore data creating meaningful information.

Our bboxtags has been designed in mind to provide a tool for image annotations that can later on be retrieved while the proper subimage can be generated dinamically by our imaging CDN.

Create an account. No credit card needed.

Create Account