AIGlobalLab specializes in custom-made AI solutions. We believe that the professional implementation of artificial intelligence technologies will solve many of the challenges facing businesses.We have developed a process that allows, on the one hand, to achieve maximum results, and, on the other hand, to optimize the cost of developing and implementing solutions.
To start a project, just send us an example of data (Table 1) and a description of the business goals that you want to achieve. We will analyze the information and offer solutions. Thus, we begin development at the presale stage.
Development process
We use a flexible development process that allows you to:
- Take into account changes in requirements during the project
- Maintain development in working condition
- Synchronize customer expectations
We achieve maximum involvement of customer’s specialists in the project implementation process.
Technologies
We do not invent a bicycle. Therefore, we rely on a wide range of developed algorithms, libraries, services.
Data & Text analysis
BigDL, NTLK, spaCy, Fairsec, OpenAI, GLTR, BERT, MeTA
Image processing
OpenCV, OpenVINO, PIL/Pillow, libvips, SimpleCV, Mahotas, SimpleITK, Cairo, GraphicsMagick & etc.
Objects detection and recognition
OpenCV, OpenVINO, Dlib, PyTorch, Keras, Caffe, TensorFlow, TensorFlow Lite, Torchvision, ML.NET, darknet/yolo/ImageNet
Voice recognition & audio processing
CMUSphinx, Sopare, Kaldi, VOSK, Google Voice, Microsoft Speech, Amazon Transcribe, Julius, eSpeak
Embedded solutions
We use all the main development tools for embedded platforms: Keil, Qube MX, Atmel Studio, Atolic, IAR, Atom, Arduino, Visual Studio & more & more. We use custom development tools. Our developers have experience in creating user/kernel mode modules for Linux/Unix, MacOS, Windows Desktop/Server/IoT, *RTOS, QNX, OpenWRT, Tiny Android, mobile OS & more.
Table 1
Sample recommended data set volume for evaluation | Size |
---|---|
Face detection | 1000 frames |
License plate recognition | 1000 frames |
Measuring the distance to the object | 30 frames |
Styled text generation | 10Mb |