English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

How can we collect and use a video dataset to further improve spatiotemporal3D Convolutional Neural Networks (3D CNNs)? In order to positively answer thisopen question in video recognition, we have conducted an exploration studyusing a couple of large-scale video datasets and 3D CNNs. In the early era ofdeep neural networks, 2D CNNs have been better than 3D CNNs in the context ofvideo recognition. Recent studies revealed that 3D CNNs can outperform 2D CNNstrained on a large-scale video dataset. However, we heavily rely onarchitecture exploration instead of dataset consideration. Therefore, in thepresent paper, we conduct exploration study in order to improve spatiotemporal3D CNNs as follows: (i) Recently proposed large-scale video datasets helpimprove spatiotemporal 3D CNNs in terms of video classification accuracy. Wereveal that a carefully annotated dataset (e.g., Kinetics-700) effectivelypre-trains a video representation for a video classification task. (ii) Weconfirm the relationships between #category/#instance and video classificationaccuracy. The results show that #category should initially be fixed, and then#instance is increased on a video dataset in case of dataset construction.(iii) In order to practically extend a video dataset, we simply concatenatepublicly available datasets, such as Kinetics-700 and Moments in Time (MiT)datasets. Compared with Kinetics-700 pre-training, we further enhancespatiotemporal 3D CNNs with the merged dataset, e.g., +0.9, +3.4, and +1.1 onUCF-101, HMDB-51, and ActivityNet datasets, respectively, in terms offine-tuning. (iv) In terms of recognition architecture, the Kinetics-700 andmerged dataset pre-trained models increase the recognition performance to 200layers with the Residual Network (ResNet), while the Kinetics-400 pre-trainedmodel cannot successfully optimize the 200-layer architecture.

Would Mega-scale Datasets Further Enhance Spatiotemporal 3D CNNs?