Components of CraftAR Enterprise Image Recognition
Image Recognition Web Service
Easy-to-deploy web service with high availability, enabling Image Recognition in the cloud via a simple REST API.
CWS allows you to set up your server environment and access the default Image Recognition web service on a Linux machine in minutes.
CWS is built on top of the Image Recognition Engine software library.
Image Recognition Engine
Highly optimized software library for Linux, capable of reliable recognition of objects by efficiently comparing images representing unknown scenes (query images) with very large collections of images representing known objects (reference images).
When used as an independent component, CRE can be used as a Python module, or included in your own C/C++ projects.
Optional component that you can use to combine CraftAR Enterprise with the CraftAR Augmented Reality SDK when developing your mobile apps.
TG generates image tracking data that is later delivered to the CraftAR Augmented Reality SDK upon recognition.
Used when running image recognition in the cloud and you also need to render interactive Augmented Reality experiences on top of the recognized objects. The AR SDK (combined with CWS) enables delivery and rendering of AR experiences for thousands of objects.
Deploy image recognition in a convenient and flexible way
Our image recognition software enables easy integration and makes use of cost-effective server infrastructure.
Pick the most convenient option when integrating image recognition with other cloud components.
CRE is delivered along with a web service code that is easy to install and immediately exposes CRE functionality via a RESTful HTTP API.
We also provide Python bindings that allows access to CRE’s functionality using Python language. With the bindings you can quickly build your own unique web services on top of the Image Recognition software.
Finally, we also provide access to the low level API in C/C++, in case you need to tightly integrate our Image Recognition software with your C/C++ projects.
The following describes an example when deploying CRE and CWS with a database of 100,000 images:
CPU requirements: Single virtual CPU core from a mid-sized off-the-shelf cloud server, e.g. Intel® Xeon E5-2680 v2 (Ivy Bridge).
RAM requirements: 2 GB for the CRE. Recommended: 2.5 GB to run the web-service and other background processes. For larger databases and a single instance of the CRE, the amount of used RAM grows linearly.
High performance image recognition at your fingertips
Our Image Recognition software provides high speed recognition and optimized image management.
Fast round trip time
The following example estimates the round-trip time for an app running on a smartphone, connected to your servers over the Internet. We include the following steps:
Sending an image of 10-15 KB to your webserver in the Internet,
CRE processing (feature extraction and search),
Returning a simple answer (e.g. JSON of 1KB) to the user device.
Those three steps (depending on network conditions) take approximately:
With Wi-Fi and 4G networks: 0,5 seconds
With 3G network: 1 to 2 seconds
Adding images in no time
One of the specific advanced features of our Image Recognition software is adding and removing images from the database in real time while your server is up and running in production.
Our technology is capable of adding a reference image and making it available for search instantaneously (takes < 0.1 second).
In practice, this means that you don’t need to be crunching visual properties or doing extensive training before an image is ready for recognition.
Contact us to learn more about our On-Premise versus Software as a Service license options.