We have developed a retina like preprocessing system where the algorithm can select interesting areas, possible object candidates for the object recognition and classifier algorithm. Our method is based on various channels encoding different spatio-temporal features of the input flows similarly to the human vision system. All these local feature calculations can be implemented on cellular neural networks and oscillatory networks this way they can be efficiently implemented on such devices.
Projektszám: 12105-401787-BS
Cím: Darpa projekt 1. fázis
Futamidő: 2014.05.21-2015.12.31.
Témavezető: Dr. Roska Tamás, Dr. Horváth András
Elnyert támogatás: 66.400,- USD
Tasks:
1. develop a software for spatiotemporal filter object detection from video framcs
2. software for object detection using oscillator models
3. software for performance analysis of itmes 1 and 2 for energy and throughput
