COMPUTER VISION
Computer vision is growing rapidly. Smartphones cameras can detect postures during fitness exercise, surveillance cameras can “read” license plate. Does your product need to “see” the world?
Computer Vision development
- Collect requirements.
- Select cameras and sensors
- Select software stack, libraries (like OpenCV), and Neural Networks
- Create Computer Vision algorythms for object detection
- Implement Machine Learning networks to automatically adapt for changing environment
- Apply Deep Learning to train Neural Networks for new objects
- Deploy. Observe. Adjust
Each project has unique set of requirements. We select sensors required to reliably capture visual information
- 2D cameras
- 3D stereoscopic cameras - Intel RealSense, Luxonis
- Time of Flight (ToF) cameras
- Lidars
- Thermal Cameras
Computer Vision is best for applications where machine need to detect pre-determined patterns and objects. Libraries like OpenCV work well for such applications as
- Object sorting
- Defect detection
- Surveilance
- Medical diagnostics
Machine Learning (ML) accompanies traditional Computer Vision to train Neural Networks to adapt to variations in observed objects or operation conditions. Neural Networks in ML algorythms go through training and afterwards may be used to detect wider range of possible scenarios
- Fitness and Sport training
- Automated surveilance systems
- Object sorting and labeling
- Advanced medical applications