npj Robotics

Table 3 Summary of generalisation performance of the three classifiers trained with O = {5, 4, 3, 2} to new grasp poses (PT and RT datasets)

From: Let’s DENSE: a novel protocol for efficiently collecting dense and diverse data for tactile slip detection in robotic grasping

		Random Forest
		Dense-Dataset		Baseline-4		Baseline-6
Training Objects		PT	RT	PT	RT	PT	RT
5	Avr.	0.40	0.51	0.32	0.31	0.31	0.36
5	std	0.04	0.07	0.06	0.07	0.09	0.14
4	Avr.	0.42	0.54	0.32	0.30	0.30	0.36
4	std	0.07	0.10	0.09	0.14	0.10	0.15
3	Avr.	0.38	0.57	0.28	0.35	0.26	0.37
3	std	0.08	0.14	0.12	0.10	0.14	0.14
2	Avr.	0.34	0.47	0.24	0.30	0.27	0.32
2	std	0.14	0.28	0.17	0.23	0.17	0.25

		Support Vector Machine
		Dense-Dataset		Baseline-4		Baseline-6
Training Objects		PT	RT	PT	RT	PT	RT
5	Avr.	0.52	0.48	0.49	0.43	0.48	0.45
5	std	0.08	0.08	0.07	0.06	0.07	0.09
4	Avr.	0.53	0.53	0.48	0.43	0.45	0.45
4	std	0.06	0.09	0.09	0.11	0.09	0.09
3	Avr.	0.49	0.51	0.42	0.43	0.42	0.43
3	std	0.12	0.13	0.14	0.11	0.14	0.13
2	Avr.	0.43	0.45	0.34	0.31	0.35	0.35
2	std	0.17	0.24	0.15	0.18	0.16	0.19

		Multi Layer Perceptron
		Dense-Dataset		Baseline-4		Baseline-6
Training Objects		PT	RT	PT	RT	PT	RT
5	Avr.	0.46	0.44	0.27	0.29	0.33	0.41
5	std	0.08	0.09	0.13	0.09	0.10	0.12
4	Avr.	0.44	0.48	0.24	0.26	0.30	0.38
4	std	0.10	0.13	0.12	0.15	0.12	0.10
3	Avr.	0.35	0.45	0.15	0.21	0.23	0.35
3	std	0.14	0.16	0.18	0.17	0.13	0.16
2	Avr.	0.31	0.42	0.15	0.19	0.14	0.24
2	std	0.15	0.22	0.17	0.19	0.16	0.20

The average and standard deviation of MCCs are computed across three repetitions and all tested subsets for each data collection approach. The best results obtained for each model and number of objects used for training are highlighted in bold.

Back to article page

Search

Advanced search

Quick links