Preloader

OUR CAPABILITIES

Solutions

Solutions

Deep Learning
for Computer Vision

Deep learning added a huge boost to the already rapidly developing field of computer vision.

With deep learning, a lot of new applications of computer vision techniques have been introduced and are now becoming parts of our everyday lives. These include:

Object Classification

Object classification involves assigning a label to an entire object or photograph.

It is a more advanced version of Image Detection. In Object Classification the neural network has to process different images with different objects, detect them and classify by the type of the item on the picture.

Object Classification

Object Detection

Object detection is the task of image classification with localization.  Any one image may contain multiple objects of different types each of which requires localization and classification. This multiplicity of objects makes this a far more challenging task than simple image classification.

At Enigma Pattern we apply various techniques for image classification with localization to provide the best possible results.

object detection

Some examples of object detection include:

  • Drawing a bounding box and labeling each object in a street scene.
  • Drawing a bounding box and labeling each object in an indoor photograph.
  • Drawing a bounding box and labeling each object in a landscape.

IMAGE STYLE TRANSFER

At Enigma Pattern we use various techniques such as style transfer, neural style transfer and GAN in the task of learning the style from one or more images and applying that style to a new image.

This task can be thought of as a type of photo filter or transformation which may not have an objective evaluation.

Examples include the application of the style of specific famous artworks (e.g. by Pablo Picasso or Vincent van Gogh) to new photographs.

Image Transformation

Image Reconstruction

Image reconstruction and image inpainting is the task of filling in missing or corrupt parts of an image.

This task can be thought of as a type of photo filter or transformation that may not have an objective evaluation.

Examples include reconstructing old, damaged black and white photographs and movies.

We often use datasets comprising existing similar photos. During the training phase of the process we create corrupted versions of existing photos that models then learn to repair.

Image Reconstruction

Image Super-Resolution

Image super-resolution is the traditionally impossible task of generating a new version of an image with higher resolution and detail than the original image.

Frequently models developed for image super-resolution can be used for image restoration and inpainting as they solve related problems.

Datasets often involve using existing photo datasets and creating down-scaled versions of photos for which the models learn to create super-resolution versions.

Image Super Resolution

Solutions

GENERAL PURPOSE
Machine Learning

The first step for any organization in the area of Big Data is to gain an understanding of the data they possess. The next is to develop insights into the possibilities of what that data can reveal.

The sheer volume and complexity of the data now being collected pushes its analysis and interpretation beyond human capability. Machine Learning is key in overcoming these limitations.  To maximise the help we give to our customers Enigma Pattern has invested heavily in developing Machine Learning capabilities.

Fraud detection

searching not only for frauds, but also other undesired user behavior. Implementation of a special background watchdog constantly analyzing every user action, and spotting suspicious behavior. Such an approach allows actions to be taken ranging from simply informing the IT team of an incident, to immediately stopping the app. The watchdog uses an AI trained model built on a sample number of people with the undesired activity, thereby learning to detect similar situations.

Users’ behavior modeling

this technique allows questions to be asked regarding actual and possible users’ behavior:

  • The best way of making the app adoption easier.
  • The most effective way to target specific customer groups such as families, elderly people, singles etc.
  • Best functionality to attract people living in selected geographic areas.
  • Predicting issues with payments.
  • Predicting the likelihood of claims of a refund.

Entities modeling

analysis of differences between versatile market entities and model actions (sales, frauds, issues, etc.) This approach leads to finding out:

  • If some entities (like shops, banks etc) are more prone to fraud than others.
  • Which situations lead to issues.
  • How to reduce unwelcome situations.

Anomalies detection

this technique detects anything that stands out from normal application use. It can find users who are trying to hack/break the application, or people who want to commit a fraud using a method that is not yet known to the customer.

Segmentation

automatic selection of coherent groups of clients and identification of common areas. Segmentation helps to determine conditions that would persuade a user to change a group eg. from regular to Heavy User.

Who has the data

has the power

Tim O'Reilly
How can we help you?

To find out more about Enigma Pattern, or to discuss how we may be of service to you,
please get in touch.

GET IN TOUCH