New SWITi method reduces neural network tiling artifacts in large image predictions

Researchers have developed SWITi, a novel test-time method designed to mitigate artifacts commonly found in tiled predictions from neural networks, particularly those used for large image data. This technique averages overlapping sliding-window predictions to smooth out discrepancies at tile seams, thereby improving reconstruction fidelity and resolution. SWITi also introduces two new metrics, the Fraction of Rejected Tests (FRT) and Artifact Severity (ASV), to detect and quantify these tiling artifacts. The method has demonstrated substantial attenuation of stitching seams and enhanced image quality on various fluorescence microscopy datasets, which is especially beneficial for downstream processing in biomedical applications. AI

IMPACT Improves accuracy and fidelity in large-scale image processing tasks, particularly relevant for biomedical data analysis.

RANK_REASON The cluster contains a research paper detailing a new method for image processing with neural networks. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Federico Carrara
• Gotit.pub
• Hugging Face
• ScienceCast

• paper
• infra

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