nyris engine delivers results below 1 second

Your request image passes several stages of recognition. You receive a unified, merged result in less than a second. Below you can find the different image recognition technologies we use on our unique platform.

Keypoints & Descriptors

Although they have been around for decades, keypoint & descriptor based algorithms deliver fast and efficient way to identify many products like packaged groceries, ad campaigns or catalog pages. They are a great replacement for QR codes.

General Neural Networks

There are many pre-trained networks available for general classification easily available trough e.g. Google Vision API, Microsoft Azure or IBM Watson. Those are great to get a general idea of the content of your image, is it a chair, a shirt, an apple or something else?

Custom Neural Networks

Sometimes you need to dig deeper into network architecture to make visual product search work for you. Our deep learning scientists are eager to develop a network that solves your challenges using an efficient training pipeline.

Hashing

The hash value of an image is unique. We extract different hash values (MD5, SHA256, Perceptual) and store them while we process your image data. This helps us find and group similar product images in a fast and efficient way.

OCR

Any text information on the submitted image can be extracted and used for search like the serial number of an asset, the vintage of a vine or license plates. Adding OCR to the pipeline easily solves many different vision use cases.

Logo Extraction

The unique shape of a logo can be easily recognized in an image and this information can be used to refine the results from previous stages.

Product & Image Data Management

All we need to start is your product or object data feed. CSV, XML, TXT, JSON – we don’t care, we can import nearly every structured format you can provide. We also optimize and augment your images to improve image recognition while importing your data.

Human in the loop

Visual data is not always available to create a robust, high precision recognition solution. In case a few minutes search time are acceptable for your use case, our human in the loop process is the prefect solution to deliver value to your customers and at the same time collect annotated data for future development.

Ease of Integration

You need specific attributes in your JSON response. You need different API-Keys for different Client-Apps? Different search indexes for all your different websites. Like us to extract specific values from the OCR Stage? We are all set up to customize your solution to your needs in no time.

schedule a demo today!