RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations
The paper introduces RaPD, a novel method for resolution-agnostic pixel diffusion in generative models. It utilizes a continuous Neural Image Field to enhance image generation quality and scalability. The approach allows for rendering at arbitrary resolutions while maintaining fixed diffusion costs.
- ▪RaPD stands for Resolution-agnostic Pixel Diffusion and addresses limitations in current generative models.
- ▪The method employs Semantic Representation Guidance for improved latent learning and a Coordinate-Queried Attention Renderer for effective rendering.
- ▪Experiments indicate that RaPD achieves superior generation quality and resolution scalability compared to previous methods.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computer Vision and Pattern Recognition arXiv:2605.15908 (cs) [Submitted on 15 May 2026] Title:RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations Authors:Yanhao Ge, Shanyan Guan, Weihao Wang, Ying Tai, Mingyu You View a PDF of the paper titled RaPD: Resolution-Agnostic Pixel Diffusion via Semantics-Enriched Implicit Representations, by Yanhao Ge and 4 other authors View PDF HTML (experimental) Abstract:Natural images are continuous, yet most generative models synthesize them on discrete grids, limiting resolution-flexible generation.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
