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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Following the recent success of deep neural networks (DNN) on video computervision tasks, performing DNN inferences on videos that originate from mobiledevices has gained practical significance. As such, previous approachesdeveloped methods to offload DNN inference computations for images to cloudservers to manage the resource constraints of mobile devices. However, when itcomes to video data, communicating information of every frame consumesexcessive network bandwidth and renders the entire system susceptible toadverse network conditions such as congestion. Thus, in this work, we seek toexploit the temporal coherence between nearby frames of a video stream tomitigate network pressure. That is, we propose ShadowTutor, a distributed videoDNN inference framework that reduces the number of network transmissionsthrough intermittent knowledge distillation to a student model. Moreover, weupdate only a subset of the student's parameters, which we call partialdistillation, to reduce the data size of each network transmission.Specifically, the server runs a large and general teacher model, and the mobiledevice only runs an extremely small but specialized student model. On sparselyselected key frames, the server partially trains the student model by targetingthe teacher's response and sends the updated part to the mobile device. Weinvestigate the effectiveness of ShadowTutor with HD video semanticsegmentation. Evaluations show that network data transfer is reduced by 95% onaverage. Moreover, the throughput of the system is improved by over three timesand shows robustness to changes in network bandwidth.

ShadowTutor: Distributed Partial Distillation for Mobile Video DNN Inference