Graded convolutional networks offer scale invariance and blur robustness

Researchers have developed graded convolutional networks that achieve scale invariance without explicit scale normalization, offering improved robustness to unknown blur. These networks utilize a retraction onto a weighted sphere of the grading, which separates image families that traditional radial normalizations cannot. Experiments on the MNIST database showed that these networks maintained high accuracy even under significant blur, outperforming networks relying on scale-normalized derivatives. AI

IMPACT Introduces a new architecture for image processing that enhances scale invariance and robustness to blur, potentially improving performance in computer vision tasks.

RANK_REASON The cluster contains a research paper detailing a novel network architecture. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• arXiv
• CORE Recommender
• Graded convolutional networks
• Hugging Face
• Lindeberg's Gaussian derivative network
• MNIST database

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →