DEEP LEARNING
Improvis Team is widely experienced in the development and deployment state-of-the-art Deep Learning algorithms, their optimization and integration into different platforms and devices. We also do research in the field of improvement the accuracy of Deep Learning algorithms using synthetic data generation. Our solutions are optimized to work on GPGPU-based devices and systems, where we use the CUDA platform. The solutions we provide are superfast, extra efficient, high-accurate, and reliable.
Artificial Intelligence Driven Robot for Citrus Harvesting
Deep Learning based real-time orange detection, recognition, targeting and tracking with locating of orange 3D coordinates within 0.5 cm accuracy.
Software System for Sport Event Capturing and Analysis
C++/OpenCV/CUDA/Deep Learning based software system to track and simulate cameraman actions during sport events, and provide analytics.
Synthetic Data Generation Flow for Gun Detection
Unity/c++/c#/OpenCV/Cloud Computing based system to generate gun images for instance segmentation algorithms to increase the accuracy and reliability of guns detection neural networks.
Synthetic Data Generation Platform for Deep Learning
Unity/c++/c#/OpenCV/Cloud Computing based system to generate synthetic images for instance segmentation algorithms to increase the accuracy and reliability of deep neural networks.
Software for Smoke and Fire Detection
Deep Learning based software for smoke and fire detection in buildings, urban and rural areas.
Software for Human Activity Detection
Deep Learning based software system to grab video streams and detect policy violations on certain areas that were pre-annotated and registered in the system.
Application for Motion Detection and Extraction
Deep Learning based software system for motion detection and extraction using dynamically updated background and optical flow.
Software System for Chip and Pins Segmentation
C++ based software system that determines the position of the chips, provides the number and length of pins, and identifies defects.
Safety Zone
C++, Python based software to detect people and equipment in real-time based on the streams received from the camera. It calculates the distance between each detected person and the danger zone and delivers an alert.