Custom AI solutions

"Forty percent of ‘AI startups’ in Europe don’t actually use AI, claims report"
The Verge

We work in the field of:

Image & Audio processing

Object Detection & Recognition

Data & Text Analysis

Embedded Solutions

we know how
Big Data

When big data utilizes machine learning, a business thrives. We can instantly retrieve vast volume of necessary information from huge heterogeneous data sources, sometimes even without human intervention. Whether it is a data stream full of variables, unstructured data, or even data based on human emotions and feelings.

AI for embedded

Microprocessors and algorithms development enables use of AI even on the cheapest devices. Embedded AI based devices can be used by service robots to perform assessment and control functions depending on their tasks. Not only would the development make service robots faster but also servers and companion mechanisms interacting with a human.

Image processing

Creation of machine learning models ensuring accurate objects identifying and recognition still remains a major problem of computer technologies. Yet, contemporary open source platforms allow easily develop, teach, and deploy models for objects detecting and recognizing. Therefor we provide a wide range of solutions from cloud platforms to microcontrollers.

Voice recognition

Speech recognition technologies serve to recognize words of speech or recordings in real time to convert it to text. Moreover, speech recognition enables identifying people by their voices. Voice recognition technologies can be used by customer services as well as in such industries as automotive, retail, security, smart devices, IIoT, and others.

Embedded

Raspberry Pi speech recognition. Onboard, without any online services like Azure, AWS, Google & etc.

Recognition of QR, barcode and other codes. Raspberry Pi, Intel Neural Compute Stick is optional.

Recognition of QR codes with ESP32 and OV7725 camera.

We do not keep or use your data

Online demos

Face detectionFace recognition is one of the basic tasks of image processing. There is a plenty of algorithms to recognize faces: Haar cascade, Histogram of Oriented Gradients, Neural Networks. Just upload a photo with one or several faces. File size can not exceed 700K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
DLib


Cat detectionRecognizing a cat is more difficult than a human face, therefore, two algorithms based on several images handling are used. File size can not exceed 700KB. We are working on the task and estimating the best results soon. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
OpenCV


QR codes recognitionThese days, you can see QR codes almost everywhere. To solve the QR code recognition problem, the recognition algorithm based on image processing is put forward. The whole process including image tilt correction, image orientation and image geometric correction allows images collected on different conditions to be quickly identified. File size can not exceed 700K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
OpenCV, ZBar


Lexical analysisLexical analysis serves to research structure of words, sentences, and texts. Furthermore, the analysis enables comparing, processing, and even creating texts in natural languages by means of AI technologies. Upload any text in English and enjoy the result. The diagram would show the similarities between the sentences. File size can not exceed 1KB. Technologies

Programming languages: Python
NLTK


Traffic analyzingMachine learning technology adds value by improving traffic monitoring and traffic data analysis. The technology enhances accuracy of data harvested from cameras and sensors causing precise recognition of certain traffic situations. Then the data can serve to either predict and prevent accidents and injuries, or reduce the peak hours traffic jams, or improve business performance, or contribute to transport infrastructure planning. File size cannot exceed 700K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
OpenCV


Objects classificationClassification of objects in the images is one of the most urgent tasks of AI. It is necessary for automotives, shops, service centers, etc. For demo, we use a simple classifier, otherwise the calculations will take a lot of time. When using the GPU, the quality and speed of recognition increases significantly. But you can try to upload image with figures of people, animals, cars, objects and evaluate the result. File size can not exceed 300K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
darknet/yolo


Objects classificationObjects classifier for the traffic recognition. More accurate and unfortunately slower. Just upload a photo with one or several cars, bicycles and etc. File size can not exceed 300K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
darknet/yolo


License plate recognitionThe range of neural networks applications is very wide. Including OCR (optical character recognition). In this example, numbers and characters are recognized on car license plates from different countries. File size can not exceed 300K. Technologies Test dataset

File type: JPEG, PNG, GIF.
Image Example:
Programming languages: C/C++
darknet/yolo


Text GeneratorJust click the “Generate Text” button and enjoy Ernest Hemingway-style text. We used “The old man and the sea” novel for teach our NN. Technologies

Programming languages: Python
Tensorflow, Tensorpack
Thanks to: Yuxin Wu

Text rewriterThis is not a commercial bot for articles spinning. But you can upload a English text file and get the modified text. File size can not exceed 1KB. Technologies

Programming languages: Python
NLTK
Thanks to: Thiago Cassimiro


new in digest

Subscribe to our Digest