TEMPL: A Resource-Efficient and Transparent Model-Free Framework for Pseudo-Label Generation in Material CT Scans

Key points

• The scarcity of large annotated datasets limits deep learning (DL) segmentation in material CT analysis.1
• It combines thresholding, morphological dilation, and a custom breadth-first search region-growing algorithm to produce polygonal annotations without pre-trained weights, labeled data, or a GPU.1
• We compare it against the zero-shot Segment Anything Model (SAM), a fully supervised 3D U-Net, and classical baselines.1
• These results support our transparent approach as a tool for annotation bootstrapping and microstructural quantification in resource-limited settings.1

Sources (1)

• [1]TEMPL: A Resource-Efficient and Transparent Model-Free Framework for Pseudo-Label Generation in Material CT Scans
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 10:48 PM

  We introduce a multi-stage, model-free framework that generates pseudo-labels directly from grayscale CT scans.

  The scarcity of large annotated datasets limits deep learning (DL) segmentation in material CT analysis.

Extractive summary: sentences quoted from the sources.