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/

Zendy Plus

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

Zendy Tools

Zendy Open

Downsampling is widely adopted to achieve a good trade-off between accuracyand latency for visual recognition. Unfortunately, the commonly used poolinglayers are not learned, and thus cannot preserve important information. Asanother dimension reduction method, adaptive sampling weights and processesregions that are relevant to the task, and is thus able to better preserveuseful information. However, the use of adaptive sampling has been limited tocertain layers. In this paper, we show that using adaptive sampling in thebuilding blocks of a deep neural network can improve its efficiency. Inparticular, we propose SSBNet which is built by inserting sampling layersrepeatedly into existing networks like ResNet. Experiment results show that theproposed SSBNet can achieve competitive image classification and objectdetection performance on ImageNet and COCO datasets. For example, theSSB-ResNet-RS-200 achieved 82.6% accuracy on ImageNet dataset, which is 0.6%higher than the baseline ResNet-RS-152 with a similar complexity. Visualizationshows the advantage of SSBNet in allowing different layers to focus ondifferent positions, and ablation studies further validate the advantage ofadaptive sampling over uniform methods.

SSBNet: Improving Visual Recognition Efficiency by Adaptive Sampling