Computer Vision and Augmented Reality
Computer Vision (CV) and Augmented Reality (AR) are cutting edge technologies that are rapidly becoming a game changer in many domains since they are enabling transformation of the human machine interactions. This is ultimately bringing new possibilities unachievable before. For example, AR technology allows to diagnose faults and maintain industrial machinery by integrating a physical shape of the device with a digital model in real-time. This approach significantly is reducing human errors and time of the process. Another applications are infotainment, people flow management, navigation, quality control, etc. Being profound experts in embedded systems we’re offering efficient CNN/DNN portation to constrained devices. We know how to achieve best performance with technology stacks like TensorFlow and OpenVINO while reducing complexity of CNN/DNN. Moreover we have a track record of combining edge data processing with cloud management in order to arrange self trained IoT system with optimized data traffic.
RTSoft has rich experience in implementing СV and AR algorithms for its customers.
• Deep Neural Networks
• Computer Vision
• OpenCV, OpenCI, CUDA
• Augmented Reality
• Unity, Unreal Engine
Having many facilities with complex industrial equipment require a large workforce trained enough to handle many troublesome situations. However, training this large workforce might not be financially viable and reading high volumes of manuals is not efficient anymore for many troublesome situations. This leads to an unintentional misuse of devices and big damage or loss in production environments.
AR allows to overcome this situation by offering FAEs remote assistance in the field. For example, displaying corresponding manuals, AR guided smart instructions, displaying maintenance check list, etc. This can reduce errors, increase efficiency and control over your workforce.
- AR-based live support video calling
- Remote AR
Algorithms from Machine Learning, Computer Vision and Augmented Reality can be implemented in HMI together with Advanced Driver-Assistance Systems (ADAS) to increase car safety and more generally road safety.
- Object detection (pedestrians, vehicles, traffic signs, traffic lights, Points of Interest)
- Gesture control for HMI
- Driver monitoring
- Safety alerts, navigation elements, infotainment features on Head-Up Display
Modern warehouses represent highly sophisticated facilities resembling busy downtowns of large cities. The efficiency of a warehouse’s loader in finding fastest routes to shelves with products, avoiding errors and clogs may increase productivity operations. Real-time tracking of current status of all operations, heat-maps of all routes in warehouse, real-time notification of ending stocks - all of these comprise a smart warehouse. Functionality:
- usage with smart glasses (ex., EPSON-BT300);
- real-time navigation just with a monocular camera (no additional dongles, QR codes, etc.);
- integration with WMS platform.
People flow analysis
Through pairing CV, DNN and IoT technologies we can build a system capable to count people indoor and outdoor, analyse people flow and foot traffic, recognise people for access management purposes. The software system and recognition algorithms can be flexibly distributed among onsite server and cloud while being configured and update totally remotely.
- People counting
- Queue counting
- Crowd counting
- Visitor counting
- Foot traffic analysis
- Face recognition
One of our customers, a CCTV system integration company, requested us to develop a system to count people and measure people flow in underground metro. The purpose of the system is to assist platform load and passenger queues. The video surveillance system had been already deployed there and RTSoft should propose a hardware and software add-on by a very competitive price.
RTSoft helps customers to build AI + IoT applications through integration of computer vision, data analysis and augmented reality. Our experience in embedded platforms (Intel Movidius, Nvidia Jetson, Google Coral) and dedicated ML frameworks (Openvino, TensorRT, TFLite) made it possible to fulfill this project.
The solution should allow an automatic rollback in case of update failures such as connectivity problem or unexpected power loss, implement versioning and secure delivery mechanisms, and be customizable for future integrations.
Additionally, the solution should include a cloud back-end service for device and software repository, content delivery and rollout management.
The target devices’s characteristics are:
- SoC: Mediatek MT7688
- Flash memory: 64MB
- Operating system: Linux (OpenWRT)
RTSoft designed and developed the firmware for a smart robotics controller that makes the controller an excellent platform for educational and business uses.
The system software implements the abstraction interface for work with various sensors and actuators like ultrasonic, camera, thermo, photo, encoder motor, servo motor and many others in Real-Time. The programming API is available in python and C/C++.
The robotics platform features the color touch display with the graphical interface, which can also be customized by a user via the programming API.
Besides on-board GUI the solution implements two remote interfaces: WebDAV and ssh. Integration with a cloud is supported via MQTT protocol.
The connectivity is available via WiFi, Bluetooth and USB Ethernet. SD card and USB flash drives can be used to distribute data to/from the controller. Multiple controllers can be connected with each other via CAN interface where a Master controller can seamlessly operate with remote sensors and actuators connected to a Slave controller(s) in the same real-time manner.
The solution was presented at the CES 2018 Show in Las Vegas.