Researchers have developed Pix2Pix-Hybrid (P2P-H), a novel conditional GAN designed to synthesize realistic crowd images for the Hajj pilgrimage, addressing the scarcity of annotated data and privacy concerns. The model utilizes a U-Net generator conditioned on multiple input channels, including structural cues and crowd density attributes, and incorporates multi-scale PatchGAN discriminators for detailed texture synthesis. This approach led to the creation of the CrowdH synthetic dataset, which, when used for training crowd-counting models, demonstrated improved performance and reduced mean absolute error compared to models trained solely on real data. AI
IMPACT Enables creation of synthetic datasets for specialized domains, improving model performance where real data is scarce.
RANK_REASON The cluster contains a research paper detailing a new model and dataset for image synthesis.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
