Fig. 6: Heterogeneous tactile force translation.
From: Training tactile sensors to learn force sensing from each other
A Signal-to-marker conversion process. The multichannel electrical signals (ii) from a sensor array (i) are unified into marker images (iii) accommodating sensors with either 3-axis sensing capability or z-axis-only sensing capability. B Representative examples of tactile images generated after M2M translation. C Histogram of \({R}^{2}\) values (i-ii) and radar plot (iii-iv) of force MAE, comparing the source-only method against the GenForce model with material compensation. D Real-time demonstration of force prediction with uSkin (3-axis) as the target domain, utilizing the GenForce model with material compensation. E Force prediction using GenForce model during dynamic contact events, qualitatively benchmarked against an ATI nano 17 F/T sensor (see Supplementary Video 2 and 3).
