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

Accuracy and processing speed are two important factors that affect the useof video object segmentation (VOS) in real applications. With the advancedtechniques of deep neural networks, the accuracy has been significantlyimproved, however, the speed is still far below the real-time needs because ofthe complicated network design, such as the requirement of the first framefine-tuning step. To overcome this limitation, we propose a novel mask transfernetwork (MTN), which can greatly boost the processing speed of VOS and alsoachieve a reasonable accuracy. The basic idea of MTN is to transfer thereference mask to the target frame via an efficient global pixel matchingstrategy. The global pixel matching between the reference frame and the targetframe is to ensure good matching results. To enhance the matching speed, weperform the matching on a downsampled feature map with 1/32 of the originalframe size. At the same time, to preserve the detailed mask information in sucha small feature map, a mask network is designed to encode the annotated maskinformation with 512 channels. Finally, an efficient feature warping method isused to transfer the encoded reference mask to the target frame. Based on thisdesign, our method avoids the fine-tuning step on the first frame and does notrely on the temporal cues and particular object categories. Therefore, it runsvery fast and can be conveniently trained only with images, as well as beingrobust to unseen objects. Experiments on the DAVIS datasets demonstrate thatMTN can achieve a speed of 37 fps, and also shows a competitive accuracy incomparison to the state-of-the-art methods.

Fast Video Object Segmentation via Mask Transfer Network