Research on Optical Soft Tactile Sensor Data Collection for Deep Learning

Optical tactile sensors have the advantages of high accuracy and small size. The measurement calibration of this kind of sensors often needs the help of deep learning. The accuracy of the dataset has a great impact on the training results of the deep learning model. We design a new method of 3D force data acquisition based on optical tactile sensor. This method solves the problem that the measured force value is deviated from the reference value after running for a long time. We use the same deep learning model in Baimukashev et al. (IEEE Robot Autom Lett 5(2):2618–2625, 2020 [1]) for comparison. The proposed method reduces the dataset error and improves the accuracy of the deep learning model.