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

Feature matters. How to train a deep network to acquire discriminativefeatures across categories and polymerized features within classes has alwaysbeen at the core of many computer vision tasks, specially for large-scalerecognition systems where test identities are unseen during training and thenumber of classes could be at million scale. In this paper, we address thisproblem based on the simple intuition that the cosine distance of features inhigh-dimensional space should be close enough within one class and far awayacross categories. To this end, we proposed the congenerous cosine (COCO)algorithm to simultaneously optimize the cosine similarity among data. Itinherits the softmax property to make inter-class features discriminative aswell as shares the idea of class centroid in metric learning. Unlike previouswork where the center is a temporal, statistical variable within one mini-batchduring training, the formulated centroid is responsible for clusteringinner-class features to enforce them polymerized around the network truncus.COCO is bundled with discriminative training and learned end-to-end with stableconvergence. Experiments on five benchmarks have been extensively conducted toverify the effectiveness of our approach on both small-scale classificationtask and large-scale human recognition problem.

Rethinking Feature Discrimination and Polymerization for Large-scale Recognition