StyleGAN-XL: Advancing Style-Based Generative Networks for Enhanced Image Synthesis with Cross-Modal Capabilities
Year: 2025
Type: conference-paper
Abstract: This research presents StyleGAN-XL, an enhanced generative model that advances style-based GAN architectures for superior image synthesis and cross-modal learning. The model introduces three innovations—Dynamic Style Modulation (DSM) for fine-grained visual control, Frequency-Aware Discrimination (FAD) for structural and textural consistency, and Cross-Modal Embedding (CME) for integrating non-visual modalities like text or audio. Experimental results on benchmark datasets such as FFHQ and CelebA-HQ demonstrate a 20% improvement in FID and 17% faster training compared to baseline models. StyleGAN-XL sets a new standard in controllable, high-fidelity, and efficient generative modeling across visual and multimodal domains.
Authors: Priyanshi Agrahari, Pushkar singh
Institution: United University
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7090828987
FWCI:
Cites:
Cited by:
Related to: