Title: | Mutual Support of Data Modalities in the Task of Sign Language Recognition |
Other Titles: | Vzájemná podpora datových modalit v úloze rozpoznávání znakového jazyka |
Authors: | Gruber, Ivan Krňoul, Zdeněk Hrúz, Marek Kanis, Jakub Boháček, Matyáš |
Citation: | GRUBER, I. KRŇOUL, Z. HRÚZ, M. KANIS, J. BOHÁČEK, M. Mutual Support of Data Modalities in the Task of Sign Language Recognition. In 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). Nashville: IEEE, 2021. s. 3419-3428. ISBN: 978-1-66544-899-4 , ISSN: 2160-7508 |
Issue Date: | 2021 |
Publisher: | IEEE |
Document type: | konferenční příspěvek ConferenceObject |
URI: | 2-s2.0-85116062397 http://hdl.handle.net/11025/47308 |
ISBN: | 978-1-66544-899-4 |
ISSN: | 2160-7508 |
Keywords: | rozpoznávání znakového jazyka;konvoluční neuronové sítě |
Keywords in different language: | sign language recognition;convolutional neural networks |
Abstract in different language: | This paper presents a method for automatic sign language recognition that was utilized in the CVPR 2021 ChaLearn Challenge (RGB track). Our method is composed of several approaches combined in an ensemble scheme to perform isolated sign-gesture recognition. We combine modalities of video sample frames processed by a 3D ConvNet (I3D), with body-pose information in the form of joint locations processed by a Transformer, hand region images transformed into a semantic space, and linguistically defined locations of hands. Although the individual models perform sub-par (60% to 93% accuracy on validation data), the weighted ensemble results in 95.46% accuracy. |
Rights: | Plný text je přístupný v rámci univerzity přihlášeným uživatelům. © IEEE |
Appears in Collections: | Konferenční příspěvky / Conference Papers (KKY) Konferenční příspěvky / Conference Papers (KKY) OBD |
Files in This Item:
File | Size | Format | |
---|---|---|---|
Gruber_Mutual_Support_of_Data_Modalities_in_the_Task_of_Sign_CVPRW_2021_paper.pdf | 554,23 kB | Adobe PDF | View/Open Request a copy |
Please use this identifier to cite or link to this item:
http://hdl.handle.net/11025/47308
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.