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

Few-shot segmentation aims to train a segmentation model that can fast adaptto a novel task for which only a few annotated images are provided. Most recentmodels have adopted a prototype-based paradigm for few-shot inference. Theseapproaches may have limited generalization capacity beyond the standard 1- or5-shot settings. In this paper, we closely examine and reevaluate thefine-tuning based learning scheme that fine-tunes the classification layer of adeep segmentation network pre-trained on diverse base classes. To improve thegeneralizability of the classification layer optimized with sparsely annotatedsamples, we introduce an instance-aware data augmentation (IDA) strategy thataugments the support images based on the relative sizes of the target objects.The proposed IDA effectively increases the support set's diversity and promotesthe distribution consistency between support and query images. On the otherhand, the large visual difference between query and support images may hinderknowledge transfer and cripple the segmentation performance. To cope with thischallenge, we introduce the local consensus guided cross attention (LCCA) toalign the query feature with support features based on their dense correlation,further improving the model's generalizability to the query image. Thesignificant performance improvements on the standard few-shot segmentationbenchmarks PASCAL-$5^i$ and COCO-$20^i$ verify the efficacy of our proposedmethod.

Boosting Few-Shot Segmentation via Instance-Aware Data Augmentation and Local Consensus Guided Cross Attention