Fig. 2: Data processing.
From: Gesture recognition with Brownian reservoir computing using geometrically confined skyrmion dynamics
a Kerr-microscope frames of the device with a skyrmion inside. b Convolutional Neural network prediction to reduce noise. Skyrmion prediction in purple, contacts and defects in turquoise, background and magnetic material in yellow. The skyrmion prediction is used to simulate the expected output signal from the artificial MTJs. c Analyzed video part with one possible emulated MTJ marked as a black square. d Validation accuracy map of SVM for different input pixels. In this case for differentiating all gestures with 87 data points in time. Locations with the highest accuracy for different number of data points in time are marked. As the emulated MTJs have a diameter of \(0.5\,{{\rm{\mu }}}{{\rm{m}}}\), only the center is marked.
