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 Tools annual

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 monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Long-tail distribution is widely spread in real-world applications. Due tothe extremely small ratio of instances, tail categories often show inferioraccuracy. In this paper, we find such performance bottleneck is mainly causedby the imbalanced gradients, which can be categorized into two parts: (1)positive part, deriving from the samples of the same category, and (2) negativepart, contributed by other categories. Based on comprehensive experiments, itis also observed that the gradient ratio of accumulated positives to negativesis a good indicator to measure how balanced a category is trained. Inspired bythis, we come up with a gradient-driven training mechanism to tackle thelong-tail problem: re-balancing the positive/negative gradients dynamicallyaccording to current accumulative gradients, with a unified goal of achievingbalance gradient ratios. Taking advantage of the simple and flexible gradientmechanism, we introduce a new family of gradient-driven loss functions, namelyequalization losses. We conduct extensive experiments on a wide spectrum ofvisual tasks, including two-stage/single-stage long-tailed object detection(LVIS), long-tailed image classification (ImageNet-LT, Places-LT, iNaturalist),and long-tailed semantic segmentation (ADE20K). Our method consistentlyoutperforms the baseline models, demonstrating the effectiveness andgeneralization ability of the proposed equalization losses. Codes will bereleased at https://github.com/ModelTC/United-Perception.

The Equalization Losses: Gradient-Driven Training for Long-tailed Object Recognition