Яндекс.Метрика

Optimum SmartLook Engine

The Optimum SmartLook Engine platform uses artificial intelligence technologies to analyze photo and video images. Platform-based solutions are designed for product recognition on store shelves, quality control of product assembly on a conveyor belt, preliminary diagnosis of diseases, face detection, and for other purposes.

When using mobile devices, recognition, analysis and issuance of recommendations occurs instantly, even without an Internet connection.

Basis features

  • Instant recognition of photos and videos even of poorly distinguishable objects.
  • Instant, comprehensive visual analysis of deviations from standards.
  • Instant recommendations (“do here to fix”), instant check of corrections (“was ---> done ---> became”).
  • Effective work on low-power smartphones, tablets (Android, Linux, iOS).
  • Achieving high recognition accuracy (up to 98%).

Benefits of the technology for mobile devices

  • Optimization of the neural network. Automatic Machine Learning (Auto-ML) and Learning Vector Quantization (LVQ) technology significantly reduce model size without sacrificing accuracy.
  • Multithreading. The preprocessing algorithms are developed using the OpenMP standard, without using the OpenCV or VXL libraries.
  • Developed in C ++. Uses memory quickly and efficiently, allows for almost instant image recognition using the limited resources of portable devices.
  • Pan. Automatic alignment and merging of two or more images into a panorama, elimination of duplicate objects using the most accurate algorithms (SIFT, SURF, BRISK).
  • Recognition of numbers. An innovative digit recognition method allows all visible digits to be identified, regardless of their spellings.
  • Built-in Android API. Accelerates the integration process by leveraging familiar technologies and eliminates compatibility issues.


quote

This tool has a number of unquestionable advantages. For example, the time spent by agents at the outlet reduced by 25%, and the number of visits increased. Our company has obtained a very efficient tool for collecting key data from the market.

Mikhail Zdrok, Director for integration and transformation, MARS Kazakhstan

Tasks

  • Offline recognition of images on a mobile device.
  • Recognition of images from CCTV cameras on the server.
  • Instant automatic completion of key compliance analysis documents.
  • Obtaining reliable analytical data.
  • Prompt transfer of information to analytics, forecasting and reporting systems

Key benefits

No internet connection is required for image recognition and analysis.
Low hardware load. Recognition is performed on mobile devices.
Low traffic consumption. Only model updates, system settings, transaction statistics.
Processing time 1-10 sec. Depends on the power of the mobile device.
Fast learning time. Automatic data collection for additional training and maintaining the accuracy of the system.
Offline mode. The mobile application remains fully functional.
Small size of models. Several megabytes.
Correction of data in place. The user can fix recognition errors right away.

Contact us

Your message has been sent!

The manager will contact you shortly.