2018
DOI: 10.1007/s12559-018-9568-7
|View full text |Cite
|
Sign up to set email alerts
|

End-to-End ConvNet for Tactile Recognition Using Residual Orthogonal Tiling and Pyramid Convolution Ensemble

Abstract: Background and Introduction Tactile recognition enables robots identify target objects or environments from tactile sensory readings. The recent advancement of deep learning and biological tactile sensing inspire us proposing an end-to-end architecture ROTConvPCE-mv that performs tactile recognition using residual orthogonal tiling and pyramid convolution ensemble. Methods Our approach uses stacks of raw frames and tactile flow as dual input, and incorporates the strength of multi-layer OTConvs (orthogonal til… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
5

Citation Types

0
6
0

Year Published

2018
2018
2023
2023

Publication Types

Select...
8

Relationship

0
8

Authors

Journals

citations
Cited by 11 publications
(6 citation statements)
references
References 56 publications
0
6
0
Order By: Relevance
“…A 5 × 5 pressure sensor array attached on a two-finger manipulator was used to acquire tactile frames (Zhang et al, 2018 ), each frame was then resized to a 1 × 25 vector and fed into the LSTM for feature extraction, afterwards, the extracted features at different sampling moments were assigned different weights via a self-attention module, finally, the weighted feature vectors were used for TOR. The stacks of tactile frames and tactile flow of which the computing scheme is similar to optical flow were used as dual input (Cao et al, 2018 ), and were extracted initial features by two residual orthogonal tiling convolutions (ROTConvs) branches, afterwards, the initial features were further refined by orthogonal tiling convolutions (OTConv), finally, the refined features were used to identify the object category through softmax classifier. A 28 × 50 pressure sensor array attached on a two-finger manipulator was used to acquire tactile data (Pastor et al, 2019 ), and then a 3D CNN was employed to acquire the time series features and accomplish the object recognition.…”
Section: Introductionmentioning
confidence: 99%
“…A 5 × 5 pressure sensor array attached on a two-finger manipulator was used to acquire tactile frames (Zhang et al, 2018 ), each frame was then resized to a 1 × 25 vector and fed into the LSTM for feature extraction, afterwards, the extracted features at different sampling moments were assigned different weights via a self-attention module, finally, the weighted feature vectors were used for TOR. The stacks of tactile frames and tactile flow of which the computing scheme is similar to optical flow were used as dual input (Cao et al, 2018 ), and were extracted initial features by two residual orthogonal tiling convolutions (ROTConvs) branches, afterwards, the initial features were further refined by orthogonal tiling convolutions (OTConv), finally, the refined features were used to identify the object category through softmax classifier. A 28 × 50 pressure sensor array attached on a two-finger manipulator was used to acquire tactile data (Pastor et al, 2019 ), and then a 3D CNN was employed to acquire the time series features and accomplish the object recognition.…”
Section: Introductionmentioning
confidence: 99%
“…Novel Convolutional Neural Networks (CNNs) are also acquiring excellent results in multiple applications such as visual object recognition [25]. These methods can be used for recognizing objects contacted through tactile sensors [26], [27], [28].…”
Section: Introductionmentioning
confidence: 99%
“…The main requirement behind this principle, which yields better results, is that there should be significant differences or diversity among the models. Many examples of the use of this principle in cognitive computation exist in the literature [44][45][46][47][48][49][50]. In accordance with the intrinsic hierarchy present in the data set, we will study two different scenarios: extrapolation with respect to different exercises and violinists.…”
Section: Introductionmentioning
confidence: 99%