Computer Vision Solutions
by ABTO Software

ABTO Software offers extensive technical expertise based on 10 years of research in computer vision domain.

Our computer vision department provides advanced image processing and video analysis solutions.

We are a team of PhDs and engineers ready to solve complex problems and committed to bring your vision to life.

We at ABTO Software drive the innovation and offer cutting edge technology solutions that work.

We design and engineer solutions for various industries:

Healthcare and Medicine

We develop computer vision and medical software solutions that assist healthcare transformation. Integration of machine learning methods into industry changes the whole landscape of diagnostic and prognostic accuracy. Imaging analytics as our expertise helps clinicians turn complex medical data into actionable knowledge.

Blood Cells Recognition Solution

Customer: US-based Company, healthcare industry

ABTO research and modeling experts created medical imagery solution for a US customer dealing with breast cancer treatment. The software employs image classification and blood analysis to detect pathological tissue changes. This computer vision solution works as a framework for white blood cell segmentation in microscopic blood images using digital image processing.

Business value:

  • substantially decreased computational costs and overall expenses
  • achieved a 10 times increase in the operational speed

Project challenges:

  • to increase classification speed while preserving accuracy
  • to integrate a set of systems and subsystems interacting via text file into a single high-performing solution
  • to apply block processing for large TIFF files

Key features:

  • detection of certain blood cells on a huge image obtained by a microscope
  • input image features calculation by SVM classifier
  • automated report on blood cells characteristics

Technologies and Instruments:

  • image classification, image processing, image segmentation
  • C++, MATLAB, OpenCV, Qt, Python

Security and Surveillance

Computer vision techniques are vital for security and intelligence purposes. In security analytics it is applied to scene and understanding and reconstruction, object tracking and visualization, virtual reality, input data analysis and classification, etc. Our company offers expertise in image processing and face recognition methods.

Face Detection and Recognition solution

Customer: an EU-based company, provider of brand security solutions

ABTO developed a face recognition solution allowing to extract features from person’s photos. These features are then used to build a specific watermark and put it on the photos together with additional personal information (name, surname, etc.). The resulting identity document is printed out and can be scanned in order to verify personal data and efficiently identify the individual.

Business value:

  • enhance efficiency of data acquisition in security and surveillance companies
  • high performance and 99% identification accuracy

Project challenges:

  • to eliminate image orientation fluctuations caused by instability or motion
  • to increase image processing speed and streamline algorithms
  • to develop a practical solution with the accuracy level fitting real-time applications and strict industry requirements
  • to optimize computational costs of face recognition and feature extraction algorithms
  • to develop a solution robust to illumination variations and geometric errors

Key features:

  • improved processing result through image stabilization
  • face features detection and comparison
  • landmark points identification and rotation vector calculation

Technologies and Instruments:

  • computer vision, image processing, Haar cascade, adaptive binarization, morphological operations, blobs
  • .NET, C++, MATLAB, OpenCV, Qt, SQLite, Tesseract OCR

Real Estate

Image analysis and computer vision are changing real estate business by making sense of the input data. ABTO Software engineers apply 3D reconstruction, image and video processing methods as proven mechanisms for taking decisions through meaningful data analysis, consequently looking at business in a holistic way.

3D Model Reconstruction Solution

Customer: a real estate agency, Canada

Our programming experts developed an impressive mobile application named Phantom Scanner for the Canadian customer. It allows to create a 3D model scene reconstruction of an apartment on any mobile device by means of video and image processing, line detection, feature points detection and homography matrix.

Business value:

  • VR and natural scene annotation for enterprise needs
  • real-time scene building and rendering
  • equipment cost elimination, works on regular consumer devices

Project challenges:

  • to turn sophisticated technology investigation into commercially successful project
  • to implement robust solution for Structure from motion problem
  • to accurately build wired model from 3D point cloud

Key features:

  • real-time edge detection using Canny method
  • simplification of obtained curves into corresponding line segments on image sequence
  • 3D model building

Technologies and Instruments:

  • augmented reality, structure from motion, computational geometry, iOS, computer vision, image processing, 3D scene reconstruction, mathematical analysis
  • C++, MATLAB, OpenCV

Government and Manufacturing

Active interaction between the academic researches and industry enables computer vision methodologies to interpret the environment for the benefit of humanity. We develop solutions for challenging problems in image and video analysis applicable for energy, defense, infrastructure, education and other public sectors.

Image Matching Solution

Customer: an EU company, quality control and security business

ABTO Software engineers conduct specific research of diverse image restoration and image deconvolution methods used in digital image processing applications. They are designed to improve the quality of blurred, noisy captured images for the purpose of obtaining quality high resolution pictures.

Business value:

  • cut overall expenses and time by eliminating need for manual inspection
  • provide advanced software interface with full-range functionality
  • enhance security level, delivers operational efficiencies and protects customer’s brand
  • batch processing of image sets and automatic reporting
  • computationally affordable algorithm turned into commercial Android/iOS application, no need to purchase specialized equipment

Project challenges:

  • to develop an algorithm invariant against rotation, illumination, and flip
  • to achieve the highest possible matching precision for captured images

Key features:

  • precision level of ≥ 95%
  • real-time image processing with prompt results
  • content recognition with simultaneous automatic image calibration
  • tags transfer to content publishers and providers
  • possibility to load images from the mobile resources or from an URL

Technologies and Instruments:

  • Android, iOS, image processing, information theory
  • MATLAB, C++, Python, Java/JNI, Qt, OpenCV

Our Technology Stack and Application Areas:

  • Our computer vision arsenal includes

    • 3D data acquisition and scene reconstruction
      • structure from motion (SFM)
      • simultaneous localization and mapping (SLAM)
    • Image- and video-based indexing, search and retrieval
    • Object counting
    • Pattern recognition
    • Facial detection and facial recognition
    • Pose estimation, action and gesture recognition
    • Eye tracking
    • Optical character recognition (OCR)
    • Video processing and analysis
      • motion detection
      • optical flow
      • object tracking,
      • video stabilization, etc.
    • Image processing
      • image enhancement and restoration, denoising
      • filtering, deconvolution, super resolution
      • image transformation and alignment
    • Image recognition and understanding
      • segmentation, clustering and indexing
      • image annotation and features extraction
      • image key feature extraction/detection and matching
      • object detection and recognition, category recognition
      • computational geometry
  • Machine learning techniques

    • Neural Networks and Deep Learning, Convolutional Neural Networks
    • K-means
    • ORB, SIFT
    • SVM, BoF
    • Haar cascades
    • genetic algorithms and many others
  • Computer vision application areas and examples

    • Security and surveillance
      • face recognition
      • object tracking and counting
      • automatic number plate recognition (ANPR)
    • Healthcare and medical applications
      • biometrics
      • medical image analysis
      • computer vision for sensory substitution
    • Mobile robotics
      • industrial vision systems
      • active and real time vision
      • hardware for vision and image processing
    • Mobile smart devices
    • Agriculture, natural science, government
      • aerial and satellite image analysis
      • computational imaging/photography
      • ANPR
    • Augmented, virtual and mixed reality
      • entertainment, training
    • Real estate, manufacturing, transportation and other biz
      • self-driving vehicles
      • 3d positioning, environment understanding
  • Programming languages

    • MATLAB/Octave
    • R
    • C++
    • Python
    • Java/Android
    • Objective-C
  • Libraries and platforms

    • OpenCV 2.x, 3.x
    • tesseract-ocr
    • MatConvNet
    • numpy and many others
    • Windows, Linux, Android, iOs

ABTO Software R&D team implements and delivers science intensive projects making innovations happen. We offer products that place us among world industry leaders in emerging technologies.

Object Counter

Our experts in mathematical modelling came up with an idea of development of an automated counting system which assists in detection, separation and counting of objects. It is a fast, non-intrusive solution that requires minimum additional resources. Our concept of image counting is based on slit imaging and allows counting all the typical objects with the accuracy level of 99.5%.

Object Counting Solution

Business value:

  • computationally inexpensive solution easily integrated into existing production lines
  • exceptionally quick and robust performance
  • 99.5% accuracy achieved without additional infrastructure investment

Project challenges:

  • to build an algorithm easily customizable for various business purposes
  • to develop a non-intrusive method for accurate object tracking and counting

Key features:

  • counts objects in the video stream
  • minimal hardware requirements
  • identifies the direction of the moving objects

Technologies and Instruments:

  • image processing, image segmentation, image classification
  • C++, MATLAB, OpenCV, Qt, Python

Calorie Counting Tool “OnMyPlate”

Our R&D department conducted a profound research to develop innovative mobile app that defines the real essence of healthy food on the mobile devices. This revolutionizing app prototype called OnMyPlate recognises and calculates the amount of calories in food items from images. With just a touch on mobile screen it gives a user an accurate real-time nutrition data.

Calorie Counting App “OnMyPlate”

Business value:

  • achieved real-time performance suitable for comfortable consumer usage
  • help users stay healthy by monitoring their calorie intake

Project challenges:

  • to train an accurate classification model based on an extensive food images database
  • to build a robust system with multi-view image variations
  • to implement computer vision technologies, clustering and machine learning for Android and iPhone devices

Key features:

  • 75% precision rate
  • identification of food items from consumer camera-enabled device
  • real-time calorie counting
  • simple, lean and intuitive interface
  • possibility to add and edit food composition
  • automatic updates available

Technologies and Instruments:

  • Android, iOS, image processing, machine learning, feed-forward artificial neural networks, image segmentation and clustering (k-means)
  • MATLAB, C++, Python, OpenCV, Qt

Such a good quality, paired with the promptness and such a great price-performance ratio is very rare. But most of all we appreciate to work with the great people ABTO Software and will continue to do so.

PHILIP NUSSBAUMERHead of ProductsUP-GREAT AG, Switzerland

ABTO has consistently delivered professional work on spec and on time.

Donald SchreckSenior Interactive ProducerCoActive Marketing Group, USA

Efficient solutions, outstanding service, pleasure to work with.

Matt EricksonIT Department ManagerPC Parts Unlimited, USA

ABTO’s been a fantastic partner for us over the past year. Together, we’ve delivered great projects – for our own use & for our customers. We’ve won repeat business with almost all of the clients for whom ABTO completed projects on our behalf.

(Name not disclosed due to NDA)ABTO Software’s Сustomer, Ireland

Our project manager has shown a level of dedication and work ethic that has far exceeded our expectations from an offshore development team. We are extremely happy working with him and look forward to continuing to do so over the next few months.

(Name not disclosed due to NDA)ABTO Software’s Сustomer, UK

If you see a partnership opportunity or have a challenging project Get in touch with us!