Figure 4
From: An extremely simple macroscale electronic skin realized by deep machine learning
The pressure evaluation process for 6 × 6 sectors. (a) Predicted vs measured displacement (pressure) curves for 1,000 test data points (primary test dataset) for each of the 36 sectors. The diagonal straight line in each graph stands for the optimal data status but the actual result deviated significantly from it. The red line shows the result from the polynomial function regression to model the deviation. (b) Predicted vs measured displacement for the other 500 test data points (confirmative test dataset) after the data transformation based on the polynomial regression result, showing the perfect coincidence between the predicted and measured displacement for every sector. This result shows a high precision pressure sensing with brilliant fitting parameters such as %RMSE (3.12) and MAE (0.03).
