LegalTech

keys 4

Global Ai integrator

1. AI Documents

Technologies used:

A set of algorithms for text recognition and processing based on AI technologies.Natural Language Processing

Task:

The company's contracts provide for a variety of conditions. At the same time, a significant part of the automated conditions are not taken into account and are not controlled.

Solution:

Create a model for the recognition and semantic evaluation of each contract in order to determine, distribute and fix deadlines, responsibilities, responsible persons, events, and so on that affect the fulfillment of the terms of the contracts.

Result:

The effectiveness of the execution of the terms of contracts is increased, the risk of liability is reduced, and strict control over the execution of the terms of contracts is increased.

2. OCR module

(Development of a model for recognizing and highlighting the content part of a document)

Task:

A set of monotonous documents is submitted to the system for input.It is necessary to build a system that can fill out the electronic version of these documents with high accuracy. It is required to build both a markup tool and a recognition model based on TD, OCR technologies

Solution:

A software package has been created that allows you to mark up and store this information about the document in the system, recognize the type of document and fill out its electronic form for a given type of document

Result:

– OCR: 96%

– Text Detection: 98%
– Document Classification: 99%
– Document Recognition: 94%
https://clck.ru/eSYt7

3. MediaTech

1. Demo example of blocking a target object on a video stream

Semantic detection and segmentation; Component: object detection and classification

Technologies used:

Tensor flow; OpenCV; Python; Cuda, iOS, WebTech, Mobile technologies

Task:

When solving various tasks of hiding the target object, it is necessary to solve the problem of detecting the object in the image and stylizing the ego in real time.

Solution:

The ability to use on any objects.

Quick use on a small number of items

Transfer training for a different type of facility

Segmentation of the object by the boundaries of the object

Recognition of the object's own movements (for example, to recognize the movements and actions of the object)

Recognition and blocking of unwanted content during online broadcast

Result:

Average quality of object detection in the image 95%

Average quality of object detection in the image 93%

Average quality of object trajectory tracking in the video stream 97%

Average quality of the recognized object in the video stream 89%

2. Advertising embedded in the video stream

Technologies used:

Semantic detection and segmentation; Component: object detection and classification, GAN / CycleGAN, Tensor flow; OpenCV; Python; Cuda, iOS, WebTech, Mobile technologies

Task:

We can recognize each object in the video stream, and this gives us the opportunity to integrate any media content in a post-processing format. Decision:

Result:

Increase brand awareness

Increase the attractiveness of the brand.

Increased attention to the brand

A new way of presenting information about your product.

Our special technology based on artificial intelligence analyzes the content to find the most effective advertising placements.

Brands can target an interested audience with cinematic-quality advertising campaigns on all devices.

Brands can easily scale campaigns by series, impressions and networks.

3. Animating photo and video content

Task:

Creation of universal media and entertainment content for the transformation of photos and videos of any format

Solution:

Using face sensor scanning using phone devices using Deep Face technology, which allows users to animate a still image of a face

Result:

Easy accessibility of content creation with GAN

Reduced costs for graphic changes

Realistic result in contrast to existing technologies

4. Brainy 21 Education App

Task:

To make the process of studying and memorizing theoretical material more understandable and interesting for schoolchildren and students, to increase involvement in the educational process

Solution:

Development of an online AR and VR educational platform with an explanation of educational material;

Result:

– The percentage of assimilation of information is above 90%;
– More than 95% of students were not distracted from the educational process;
– In the group with the traditional approach to explaining the material, the indicators were 2-3 times lower

Gamification of the educational process through a system of scoring and rewards

5. AR in the field of tourism

Task:

To increase the tourist attractiveness of the area and increase the tourist flow

Solution:

Develop an augmented reality mobile application in which you can enrich the real environment by adding a list of virtual audiovisual effects, such as videos, 3D models, animations and custom sounds.

Result:

A bright informational occasion; positive public opinion regarding the tourist season and professions of the tourism and hospitality industry.

6. Event recognition by cameras

Task:

Using the city video surveillance system, the relevant services should signal about illegal events (a fight, garbage thrown by a person, garbage not cleaned in the yard by housing and communal services, an accident, etc.)

Solution:

A complex of neural network algorithms trained to recognize target classes of events

Result:

Automatic processing of video footage from CCTV cameras, increasing the accuracy of event detection by 80%, increasing the response time of services by 60%

7. Video conversion to cartoon (Stylization)

Technologies used:

GAN, neural networks, style transfer, face detection, Tensorflow, PyTorch, Cuda

Task:

Creation of universal media and entertainment content for the transformation of video into a digital avant-garde/art or watercolors

Result:

Creating unique video content that causes a WOW effect

Attracting new subscribers and customers

Solution:

Digital Art,

Colored Avantgarde,

Portrait

and others

Using different styles for better and more attractive content:

8. Face transfer (DEEPFIRE technology)

Technologies used:

GAN, neural networks, face detection, deepface

Task:

Transfer of the selected face to a given video sequence (both in the video and in the photo). Generating new faces and voices

Result:

Easy accessibility of content creation with deep face technology

Reducing the cost of working with graphics, production, video editing

Realistic result of transferring faces to any video

Technical stack: Tensorflow, PyTorch, Cuda

Solution:

Based on the deep face technology and the input image of the face, the specified face is transferred to the selected video sequence to the found faces in the frames (it is possible for one, it is possible for several). The overlay occurs frame-by-frame, both in the photo and in the video.

Ready to start something great?