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

The mainstream workflow of image recognition applications is first trainingone global model on the cloud for a wide range of classes and then servingnumerous clients, each with heterogeneous images from a small subset of classesto be recognized. From the cloud-client discrepancies on the range of imageclasses, the recognition model is desired to have strong adaptiveness,intuitively by concentrating the focus on each individual client's localdynamic class subset, while incurring negligible overhead. In this work, wepropose to plug a new intra-client and inter-image attention (ICIIA) moduleinto existing backbone recognition models, requiring only one-time cloud-basedtraining to be client-adaptive. In particular, given a target image from acertain client, ICIIA introduces multi-head self-attention to retrieve relevantimages from the client's historical unlabeled images, thereby calibrating thefocus and the recognition result. Further considering that ICIIA's overhead isdominated by linear projection, we propose partitioned linear projection withfeature shuffling for replacement and allow increasing the number of partitionsto dramatically improve efficiency without scarifying too much accuracy. Wefinally evaluate ICIIA using 3 different recognition tasks with 9 backbonemodels over 5 representative datasets. Extensive evaluation results demonstratethe effectiveness and efficiency of ICIIA. Specifically, for ImageNet-1K withthe backbone models of MobileNetV3-L and Swin-B, ICIIA can improve the testingaccuracy to 83.37% (+8.11%) and 88.86% (+5.28%), while adding only 1.62% and0.02% of FLOPs, respectively.

One-Time Model Adaptation to Heterogeneous Clients: An Intra-Client and Inter-Image Attention Design