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

Superpixel segmentation has recently seen important progress benefiting fromthe advances in differentiable deep learning. However, the very high-resolutionsuperpixel segmentation still remains challenging due to the expensive memoryand computation cost, making the current advanced superpixel networks fail toprocess. In this paper, we devise Patch Calibration Networks (PCNet), aiming toefficiently and accurately implement high-resolution superpixel segmentation.PCNet follows the principle of producing high-resolution output fromlow-resolution input for saving GPU memory and relieving computation cost. Torecall the fine details destroyed by the down-sampling operation, we propose anovel Decoupled Patch Calibration (DPC) branch for collaboratively augment themain superpixel generation branch. In particular, DPC takes a local patch fromthe high-resolution images and dynamically generates a binary mask to imposethe network to focus on region boundaries. By sharing the parameters of DPC andmain branches, the fine-detailed knowledge learned from high-resolution patcheswill be transferred to help calibrate the destroyed information. To the best ofour knowledge, we make the first attempt to consider the deep-learning-basedsuperpixel generation for high-resolution cases. To facilitate this research,we build evaluation benchmarks from two public datasets and one new constructedone, covering a wide range of diversities from fine-grained human parts tocityscapes. Extensive experiments demonstrate that our PCNet can not onlyperform favorably against the state-of-the-arts in the quantitative results butalso improve the resolution upper bound from 3K to 5K on 1080Ti GPUs.

Generating Superpixels for High-resolution Images with Decoupled Patch Calibration