From Actions to Events: A Transfer Learning Approach Using Improved Deep Belief Networks Article Swipe

MAKE IMPACT

Mateus Roder , Jurandy Almeida , Gustavo Henrique de Rosa , Leandro A. Passos , André Luis Debiaso Rossi , João Paulo Papa · YOU? · · 2021 · Open Access · · DOI: https://doi.org/10.1109/ssci50451.2021.9660128

In the last decade, exponential data growth supplied machine learning-based\nalgorithms' capacity and enabled their usage in daily-life activities.\nAdditionally, such an improvement is partially explained due to the advent of\ndeep learning techniques, i.e., stacks of simple architectures that end up in\nmore complex models. Although both factors produce outstanding results, they\nalso pose drawbacks regarding the learning process as training complex models\nover large datasets are expensive and time-consuming. Such a problem is even\nmore evident when dealing with video analysis. Some works have considered\ntransfer learning or domain adaptation, i.e., approaches that map the knowledge\nfrom one domain to another, to ease the training burden, yet most of them\noperate over individual or small blocks of frames. This paper proposes a novel\napproach to map the knowledge from action recognition to event recognition\nusing an energy-based model, denoted as Spectral Deep Belief Network. Such a\nmodel can process all frames simultaneously, carrying spatial and temporal\ninformation through the learning process. The experimental results conducted\nover two public video dataset, the HMDB-51 and the UCF-101, depict the\neffectiveness of the proposed model and its reduced computational burden when\ncompared to traditional energy-based models, such as Restricted Boltzmann\nMachines and Deep Belief Networks.\n

Type

article

Language

en

Landing Page

https://doi.org/10.1109/ssci50451.2021.9660128

OA Status

green

Cited By

2

References

33

Related Works

10

OpenAlex ID

https://openalex.org/W4206939075

Raw OpenAlex JSON

OpenAlex ID

https://openalex.org/W4206939075

Canonical identifier for this work in OpenAlex

DOI

https://doi.org/10.1109/ssci50451.2021.9660128

Digital Object Identifier

Title

From Actions to Events: A Transfer Learning Approach Using Improved Deep Belief Networks

Work title

Type

article

OpenAlex work type

Language

en

Primary language

Publication year

2021

Year of publication

Publication date

2021-12-05

Full publication date if available

Authors

Mateus Roder, Jurandy Almeida, Gustavo Henrique de Rosa, Leandro A. Passos, André Luis Debiaso Rossi, João Paulo Papa

List of authors in order

Landing page

https://doi.org/10.1109/ssci50451.2021.9660128

Publisher landing page

Open access

Yes

Whether a free full text is available

OA status

green

Open access status per OpenAlex

OA URL

https://arxiv.org/pdf/2211.17045

Direct OA link when available

Concepts

Computer science, Artificial intelligence, Machine learning, Deep learning, Transfer of learning, Process (computing), Deep belief network, Boltzmann machine, Action (physics), Adaptation (eye), Domain (mathematical analysis), Restricted Boltzmann machine, Event (particle physics), Domain knowledge, Operating system, Mathematical analysis, Physics, Optics, Mathematics, Quantum mechanics

Top concepts (fields/topics) attached by OpenAlex

Cited by

2

Total citation count in OpenAlex

Citations by year (recent)

2025: 1, 2023: 1

Per-year citation counts (last 5 years)

References (count)

33

Number of works referenced by this work

Related works (count)

10

Other works algorithmically related by OpenAlex

Full payload