Neural networks and artificial intelligence
Artificial Intelligence and soft-computing in general are particularly useful tools in areas where the knowledge-based procedures designed by domain specialists are used for decision-making. We believe that the proper use of artificial intelligence offers a unique opportunity to use existing data, photographs and camera records to designed innovative automation solutions, which improve the quality and efficiency of work.
We have an extensive experience with selecting a suitable neural network architecture for specific use cases. We use neural networks in the field of image recognition and object detection, especially in situations where the basic algorithmic solutions are inadequate, due to variability of scene conditions. We employ modified neural network implementations for object recognition (ResNet, eXception), as well as neural networks that allow detection of objects contained in the image, especially Yolo (YOLO v3) network and models derived from this architecture.
USPIN artificial intelligence : Bird eggs detection
To detect defects and noise in industrial applications, we commonly utilize neural networks to detect anomalies, especially in situations when customers do not require detailed classification of detected errors. This approach has much lower requirements for collecting and describing error situations. Images are eventually sorted into "correct" / "faulty" categories by artificial intelligence.