New framework generates pseudo-labels for material CT scans without GPUs

Researchers have developed TEMPL, a novel framework for generating pseudo-labels in material CT scans without requiring pre-trained models or GPUs. This model-free approach utilizes thresholding, morphological dilation, and a region-growing algorithm to create annotations efficiently. TEMPL demonstrates superior performance compared to the Segment Anything Model (SAM) and traditional methods in terms of accuracy and speed, offering a valuable tool for annotation bootstrapping and microstructural quantification in resource-constrained environments. AI

IMPACT Provides a resource-efficient method for generating training data for segmentation tasks, potentially accelerating research in material science.

RANK_REASON The cluster describes a new research paper detailing a novel framework for pseudo-label generation in material CT scans. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• 3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation
• arXiv
• Hugging Face
• Material CT
• Segment Anything Model
• YOLO

• paper
• other

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