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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Training robust supervised deep learning models for many geospatialapplications of computer vision is difficult due to dearth of class-balancedand diverse training data. Conversely, obtaining enough training data for manyapplications is financially prohibitive or may be infeasible, especially whenthe application involves modeling rare or extreme events. Syntheticallygenerating data (and labels) using a generative model that can sample from atarget distribution and exploit the multi-scale nature of images can be aninexpensive solution to address scarcity of labeled data. Towards this goal, wepresent a deep conditional generative model, called VAE-Info-cGAN, thatcombines a Variational Autoencoder (VAE) with a conditional InformationMaximizing Generative Adversarial Network (InfoGAN), for synthesizingsemantically rich images simultaneously conditioned on a pixel-level condition(PLC) and a macroscopic feature-level condition (FLC). Dimensionally, the PLCcan only vary in the channel dimension from the synthesized image and is meantto be a task-specific input. The FLC is modeled as an attribute vector in thelatent space of the generated image which controls the contributions of variouscharacteristic attributes germane to the target distribution. An interpretationof the attribute vector to systematically generate synthetic images by varyinga chosen binary macroscopic feature is explored. Experiments on a GPStrajectories dataset show that the proposed model can accurately generatevarious forms of spatio-temporal aggregates across different geographiclocations while conditioned only on a raster representation of the roadnetwork. The primary intended application of the VAE-Info-cGAN is syntheticdata (and label) generation for targeted data augmentation for computervision-based modeling of problems relevant to geospatial analysis and remotesensing.

VAE-Info-cGAN: Generating Synthetic Images by Combining Pixel-level and Feature-level Geospatial Conditional Inputs