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

Ashiqur Rahman; Venkata Devesh Reddy Seethi; Austin Yunker; Rajkumar Kettimuthu; Hamed Alhoori

cs.CVOct 20, 2026

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

Authors: Ashiqur Rahman, Venkata Devesh Reddy Seethi, Austin Yunker, Rajkumar Kettimuthu, Hamed Alhoori

Organizations: Department of Computer Science, Northern Illinois University, DeKalb, IL, USA · Argonne National Laboratory, Lemont, IL, USA

Abstract

The scarcity of large annotated datasets limits deep learning (DL) segmentation in material CT analysis. DL-based pseudo-labeling can help, but it is often computationally expensive and hard to interpret. We introduce a multi-stage, model-free framework that generates pseudo-labels directly from grayscale CT scans. 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. We compare it against the zero-shot Segment Anything Model (SAM), a fully supervised 3D U-Net, and classical baselines. On a dendrite dataset with ground truth, our method reaches 63.19% IoU versus 10.74% for zero-shot SAM, with four times the instance recall at 0.5 IoU (33.3% versus 8.3%). The supervised 3D U-Net serves as a reference ceiling that requires expert labels. On a silica sand specimen, YOLO models trained on our labels performed comparably to models trained on SAM labels in a qualitative comparison. Across 1,053 extracted grains, the median equivalent diameter was 163.8 micrometers, near the specimen's known 106 to 150 micrometers sieve range, compared with 219.5 micrometers for SAM, whose labels merge adjacent particles and add spurious detections. A sensitivity analysis shows a stable operating range for the key threshold and a sharp, predictable failure point linked to phase contrast. Our method needs 15s per image on a standard CPU versus 28s for SAM on a GPU. These results support our transparent approach as a tool for annotation bootstrapping and microstructural quantification in resource-limited settings.

Figures & tables

Panel of Figure 1 — Fig. 1: Representative samples from the datasets.

Fig. 2: Steps to create pseudo label from CT-scan images

Panel of Figure 3 — Fig. 3: Threshold-based label mask generated from the original image.

Panel of Figure 4 — Fig. 4: Label mask generated from the dilated outlines.

Panel of Figure 5 — Fig. 5: Masks and pseudo-labels for image segmentation models.

Panel of Figure 6 — Fig. 6: Label generation attempts from W-Net and CUTS models.

Panel of Figure 7 — Fig. 7: Comparison of object detection pseudo-labels generated using model-free approach and SAM.

Method	Parameters(Mi)	CPU	GPU	RAM	Time(sec)
Model-free	$-$	$\checkmark$	$\times$	$\checkmark$	$15/im$
SAM	$80.9$	$\checkmark$	$\checkmark$	$\checkmark$	$28/im$

Table 1: Comparison of resources for label generation showing the faster throughput of the model-free approach without GPU resources.

Panel of Figure 9 — Fig. 8: Comparison of image segmentation pseudo-labels generated using model-free approach and SAM.

Panel of Figure 10 — Fig. 9: Comparison of object detection models trained with model-free pseudo-labels and SAM-based pseudo-labels

Panel of Figure 11 — Fig. 10: Comparison of instance segmentation models trained with model-free pseudo-labels and SAM-based pseudo-labels

Panel of Figure 12 — Fig. 11: Visual comparison of pseudo-labels generated by our algorithm (c) with the Watershed approach (b), and zero-shot SAM (d).

Method	Accuracy	IoU	F1
SAM	$98.63\%$	$10.74\%$	$19.39\%$
Watershed @150px	$98.50\%$	$2.86\%$	$5.56\%$
Watershed @10px	$99.06\%$	$60.76\%$	$75.59\%$
3D U-Net [ 20 ]	$99.84\%$	$95.56\%$	$98.59\%$
Model-free	$99.12\%$	$63.19\%$	$77.44\%$

Table 2: Pixel-level quantitative comparison of pseudo-label generation showing high performance of the model-free approach with zero labeled data. The 3D U-Net shows a reference ceiling where a large set of labeled data is required.

Method	$AP@{50}$	$AR@{50}$	$AP@{75}$	$AR@{75}$
SAM	$100.00\%$	$8.33\%$	$100.00\%$	$8.33\%$
Watershed @150px	$0.00\%$	$0.00\%$	$0.00\%$	$0.00\%$
Watershed @10px	$2.63\%$	$41.67\%$	$2.63\%$	$41.67\%$
Model-free	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$

Table 3: Instance-level quantitative comparison of pseudo-label generation showing a reliable performance of the model-free framework.

Panel of Figure 15 — Fig. 12: Comparison of ECD calculated by model-free approach and SAM at zero-shot setting.

Fig. 13: Statistical comparison of grain size distributions. Left, kernel density estimates of the ECD distributions for model-free and SAM pseudo-labels. Right, corresponding empirical cumulative distribution functions, with the KS statistic corresponding to their maximum vertical separation.

Threshold	Prec.	Rec.	F1	AP@50	AR@50	AP@75	AR@75
$\tau-4$	$63.40\%$	$99.99\%$	$77.60\%$	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$
$\tau-3$	$63.40\%$	$99.99\%$	$77.60\%$	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$
$\tau-2$	$63.40\%$	$99.99\%$	$77.60\%$	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$
$\tau-1$	$63.40\%$	$99.99\%$	$77.60\%$	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$
$\tau+0$	$63.19\%$	$99.99\%$	$77.44\%$	$57.14\%$	$33.33\%$	$28.57\%$	$16.67\%$
$\tau+1$	$6.31\%$	$99.99\%$	$11.87\%$	$6.25\%$	$16.67\%$	$0.00\%$	$0.00\%$

Table 4: Framework performance across variations of the intensity similarity threshold (

\tau

Image Segmentation

Image segmentation assigns pixels or voxels to objects and regions, supporting measurement, scene understanding, and decisions across image domains.

See all 399 papers

Unsupervised Semantic Segmentation in Synchrotron Computed Tomography with Self-Correcting Pseudo Labels
Feb 27, 2026Austin Yunker, Peter Kenesei, Hemant Sharma +3Unsupervised Image Segmentation 3D Semantic Segmentation
Data Science and Learning Division, Argonne National Laboratory, Lemont, IL USA · X-Ray Science Division, Argonne National Laboratory, Lemont, IL USA
Foundation Model-Enabled Efficient Data Sampling (FEEDS): A label-efficient training strategy for pan-cancer, multi-tracer PET/CT datasets
Aug 11, 2026Biratal Raj Wagle, Bashirul Azam Biswas, Grant Chau +5Data Selection Medical Image Segmentation
Department of Biomedical Data Science Geisel School of Medicine at Dartmouth Hanover, NH 03755, USA · Radiology Dartmouth Hitchcock Medical Center Lebanon, NH 03766, USA · Department of Urology Yale University New Haven, CT, USA +1
Efficient Semi-Automated Material Microstructure Analysis Using Deep Learning: A Case Study in Additive Manufacturing
Mar 14, 2026Sanjeev S. Navaratna, Nikhil Thawari, Gunashekhar Mari +3Interactive Image Segmentation Image Segmentation
Department of Metallurgical and Materials Engineering, Indian Institute of Technology Madras, Chennai 600036, India · Center for Atomistic Modelling and Materials Design, IIT Madras, Chennai 600036, India

Feb 27, 2026J/K move · Enter open · S save

Unsupervised Semantic Segmentation in Synchrotron Computed Tomography with Self-Correcting Pseudo Labels

Austin Yunker, Peter Kenesei, Hemant Sharma, Jun-Sang Park+2

Data Science and Learning Division, Argonne National Laboratory, Lemont, IL USA · X-Ray Science Division, Argonne National Laboratory, Lemont, IL USA

Unsupervised Image Segmentation 3D Semantic Segmentation Pseudo-Labeling Computed Tomography

Open paper AskPDF arXiv

X-ray computed tomography (CT) is a widely used imaging technique that provides detailed examinations into the internal structure of an object with synchrotron CT (SR-CT) enabling improved data quality by using higher energy, monochromatic X-rays. While SR-CT allows for improved resolution, time-resolved experimentation, and reduced imaging artifacts, it also produces significantly larger datasets than conventional CT. Accurate and efficient evaluation of these datasets is a critical component of these workflows; yet is often done manually representing a major bottleneck in the analysis phase. While deep learning has emerged as a powerful tool capable of providing a wide range of purely data-driven solutions, it requires a substantial amount of labeled data for training and manual annotation of SR-CT datasets is impractical in practice. In this paper, we introduce a novel framework that enables automatic segmentation of large, high-resolution SR-CT datasets by eliminating the need to hand label images for deep learning training. First, we generate pseudo labels by clustering on the voxel values identifying regions in the volume with similar attenuation coefficients producing an initial semantic map. Afterwards, we train a segmentation model on the pseudo labels before utilizing the Unbiased Teacher approach to self-correct them ensuring accurate final segmentations. We find our approach improves pixel-wise accuracy and mIoU by 13.31% and 15.94%, respectively, over the baseline pseudo labels when using a magnesium crystal SR-CT sample. Additionally, we extensively evaluate the different components of our workflow including segmentation model, loss function, pseudo labeling strategy, and input type. Finally, we evaluate our approach on to two additional samples highlighting our frameworks ability to produce segmentations that are considerably better than the original pseudo labels.

Abstract

Image Segmentation

Explore similar work

Unsupervised Semantic Segmentation in Synchrotron Computed Tomography with Self-Correcting Pseudo Labels

Foundation Model-Enabled Efficient Data Sampling (FEEDS): A label-efficient training strategy for pan-cancer, multi-tracer PET/CT datasets

Efficient Semi-Automated Material Microstructure Analysis Using Deep Learning: A Case Study in Additive Manufacturing