ServReality

IKNOWWHATYOUMEAN/MACHINE LEARNING +AI

Product overview: IKnowWhatYouMean is a robust real-time speech recognition application, which can convert voice into text. You can talk to the computer and it will type the text in a snap.

Issues: The software had to type exactly the words spoken with great accuracy. It had to rapidly identify the language, the topic, and the context and transcribe what is being discussed in a snap. The IKnowWhatYouMean app should be able to comprehend and manage the voice from audio and from a microphone as well. Plus, we had to develop and implement meaningful analytics, so that it could structure and analyze the text. Due to machine intelligence techniques, we precisely transcribed the voice. It allowed analyzing grammar and language structure data, taking into account how the audio signal is composed. Plus, the customer wanted the app to be a self-learning creation. The IKnowWhatYouMean app should support 5 languages: English, French, German, Russian, and Spanish.

Supported audio formats: Web Media (WebM), Opus or Vorbis codec, MP3 or MPEG, Waveform Audio File Format (WAV), Linear 16-bit Pulse-Code Modulation (PCM), Free Lossless Audio Codec (FLAC), mu-law (or u-law) audio, and basic audio.

Technologies used: Application Programming Interfaces (APIs), WebSocket, HTTP REST

Result: A powerful, yet elegant and easygoing application, which can be utilized in various industries where you might need to use the voice recognition option.

VTRAINER/ VIRTUAL REALITY/ MACHINE LEARNING

Product overview: VTrainer is a smart virtual reality mobile app that lets you learn anything you want. You specify what you wish to learn, for example, to drive a car, play volleyball or tennis, and the app becomes your virtual trainer. The main idea of the app is that you are not watching someone else doing a task, or listening to how another person would do it; you are learning to do it yourself and the VTrainer helps you.

Issues: the client asked us to develop a concept app and system. The app should quickly identify the required subjects, browse through its data to find the best training solution or several ones, show them, letting users choose the best virtual training method for them. Once chosen, the app assists during the whole learning process. We needed to integrate existing voice technologies into the application by means of special APIs and other development tools. The app should be cross-platform, so that the same assistant could be used on smartphone, tablet, or laptop. Since it is a machine-learning app, gets smarter upon each use.

Technologies used: CAVE, Oculus, HTC Vive, Machine Learning

AI DRIVE/ AI AND DEEP LEARNING

Product overview: AI Drive is a smart AI-based computer for the cars, trucks and other vehicles that helps car manufacturers and auto details suppliers hasten the automated and self-driving vehicles production.

Issues: We needed to create such a machine computer that not only tracks the environment around the driver, but also perfectly realizes what’s happening around the vehicle in real-time and detects all static and dynamic objects in the area. By accurately locating itself on the HD map AI Drive had to create and then plan the best suiting and what’s more important – safe way to go. We needed a full 360-degree environment around the car. The computer had to be not only robust and feature-packed, including the weatherproof scanning option, but it should also be easy to use. Plus, it should be able to mimic the human brain’s capacity to make decisions based on a number of incidents. In that way, the self-driving vehicle would have numerous options to choose from if something unexpected occurs. Another feature we needed to implement in AI Drive is the ability to accurately communicate with the other self-driving cars around.

Technologies used: deep learning, sensor fusion, and surround vision are greatly combined to enhance the driving experience. 360° video, 3D visualization

SWEETHOMEAR/ AUGMENTED REALITY

Product overview: SweetHomeAR is an augmented reality app, which makes it possible to virtually test drive furniture and home goods. With it, you can check how any item would look like in your house, room or any other place. The camera of your phone will capture the item you wish to check, the app will analyze its characteristics (size, material, color, etc.). Then you need to specify where you wish to place this item in your house, room, flat, any other premise. The app then analyzes the surroundings and shows you how that item would look like as realistic as possible.

Issues: We needed to create photo-realistic, highly interactive visualizations of places for augmented reality. We had to properly apply the mixture of spatial presentation with real-time configuration capability. Plus, that should be really simple and engaging to use. Additionally, the app should be IOs and Android compatible.

Technologies used: Android SDK, Android NDK, C/C++