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

It is not trivial to optimally learn a 3D Convolutional Neural Networks (3DConvNets) due to high complexity and various options of the training scheme.The most common hand-tuning process starts from learning 3D ConvNets usingshort video clips and then is followed by learning long-term temporaldependency using lengthy clips, while gradually decaying the learning rate fromhigh to low as training progresses. The fact that such process comes along withseveral heuristic settings motivates the study to seek an optimal "path" toautomate the entire training. In this paper, we decompose the path into aseries of training "states" and specify the hyper-parameters, e.g., learningrate and the length of input clips, in each state. The estimation of the kneepoint on the performance-epoch curve triggers the transition from one state toanother. We perform dynamic programming over all the candidate states to planthe optimal permutation of states, i.e., optimization path. Furthermore, wedevise a new 3D ConvNets with a unique design of dual-head classifier toimprove spatial and temporal discrimination. Extensive experiments on sevenpublic video recognition benchmarks demonstrate the advantages of our proposal.With the optimization planning, our 3D ConvNets achieves superior results whencomparing to the state-of-the-art recognition methods. More remarkably, weobtain the top-1 accuracy of 80.5% and 82.7% on Kinetics-400 and Kinetics-600datasets, respectively. Source code is available athttps://github.com/ZhaofanQiu/Optimization-Planning-for-3D-ConvNets.

Optimization Planning for 3D ConvNets