Next Article in Journal
Advances in Optical Image Analysis Textural Segmentation in Ironmaking
Next Article in Special Issue
Robust Hand Shape Features for Dynamic Hand Gesture Recognition Using Multi-Level Feature LSTM
Previous Article in Journal
Salivary Redox Biomarkers in the Course of Caries and Periodontal Disease
Previous Article in Special Issue
Gesture Recognition Based on 3D Human Pose Estimation and Body Part Segmentation for RGB Data Input
Article

Learning Class-Specific Features with Class Regularization for Videos

Department of Information and Computing Sciences, Utrecht University, Princetonplein 5, 3584 CC Utrecht, The Netherlands
*
Author to whom correspondence should be addressed.
Appl. Sci. 2020, 10(18), 6241; https://doi.org/10.3390/app10186241
Received: 28 July 2020 / Revised: 2 September 2020 / Accepted: 4 September 2020 / Published: 8 September 2020
(This article belongs to the Special Issue Deep Learning-Based Action Recognition)
One of the main principles of Deep Convolutional Neural Networks (CNNs) is the extraction of useful features through a hierarchy of kernels operations. The kernels are not explicitly tailored to address specific target classes but are rather optimized as general feature extractors. Distinction between classes is typically left until the very last fully-connected layers. Consequently, variances between classes that are relatively similar are treated the same way as variations between classes that exhibit great dissimilarities. In order to directly address this problem, we introduce Class Regularization, a novel method that can regularize feature map activations based on the classes of the examples used. Essentially, we amplify or suppress activations based on an educated guess of the given class. We can apply this step to each minibatch of activation maps, at different depths in the network. We demonstrate that this improves feature search during training, leading to systematic improvement gains on the Kinetics, UCF-101, and HMDB-51 datasets. Moreover, Class Regularization establishes an explicit correlation between features and class, which makes it a perfect tool to visualize class-specific features at various network depths. View Full-Text
Keywords: class regularization; 3D-CNN; spatiotemporal activations; class-specific features class regularization; 3D-CNN; spatiotemporal activations; class-specific features
Show Figures

Figure 1

MDPI and ACS Style

Stergiou, A.; Poppe, R.; Veltkamp, R.C. Learning Class-Specific Features with Class Regularization for Videos. Appl. Sci. 2020, 10, 6241. https://doi.org/10.3390/app10186241

AMA Style

Stergiou A, Poppe R, Veltkamp RC. Learning Class-Specific Features with Class Regularization for Videos. Applied Sciences. 2020; 10(18):6241. https://doi.org/10.3390/app10186241

Chicago/Turabian Style

Stergiou, Alexandros, Ronald Poppe, and Remco C. Veltkamp. 2020. "Learning Class-Specific Features with Class Regularization for Videos" Applied Sciences 10, no. 18: 6241. https://doi.org/10.3390/app10186241

Find Other Styles
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.

Article Access Map by Country/Region

1
Back to TopTop