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

Recent advances in self-supervised learning have integrated Masked ImageModeling (MIM) and Siamese Networks into a unified framework that leverages thebenefits of both techniques. However, several issues remain unaddressed whenapplying conventional erase-based masking with Siamese ConvNets. These include(I) the inability to drop uninformative masked regions in ConvNets as theyprocess data continuously, resulting in low training efficiency compared to ViTmodels; and (II) the mismatch between erase-based masking and thecontrastive-based objective in Siamese ConvNets, which differs from the MIMapproach. In this paper, we propose a filling-based masking strategy calledMixMask to prevent information incompleteness caused by the randomly erasedregions in an image in the vanilla masking method. Furthermore, we introduce aflexible loss function design that considers the semantic distance changebetween two different mixed views to adapt the integrated architecture andprevent mismatches between the transformed input and objective in MaskedSiamese ConvNets (MSCN). We conducted extensive experiments on variousdatasets, including CIFAR-100, Tiny-ImageNet, and ImageNet-1K. The resultsdemonstrate that our proposed framework achieves superior accuracy on linearprobing, semi-supervised, and supervised finetuning, outperforming thestate-of-the-art MSCN by a significant margin. Additionally, we demonstrate thesuperiority of our approach in object detection and segmentation tasks. Oursource code is available at https://github.com/LightnessOfBeing/MixMask.

MixMask: Revisiting Masking Strategy for Siamese ConvNets