Keuper Labs
← All publications

Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVw 2025) · 2025

Smart Eyes for Silent Threats - VLMs and In-Context Learning for THz Imaging

Nicolas Poggi, Shashank Agnihotri, Margret Keuper

Vision-Language ModelsIn-Context LearningTerahertz ImagingFew-Shot Learning

Why this publication matters

Specialized terahertz images are useful in settings where large labeled datasets may be unavailable. This study gives a vision-language model a small number of relevant examples and instructions tailored to that imaging method. It explores a way to obtain useful classifications and explanations without a full retraining process.

Abstract

Terahertz (THz) imaging enables non-invasive analysis for applications such as security screening and material classification, but effective image classification remains challenging due to limited annotations, low resolution, and visual ambiguity. We introduce In-Context Learning (ICL) with Vision-Language Models (VLMs) as a flexible, interpretable alternative that requires no fine-tuning. Using a modalityaligned prompting framework, we adapt two open-weight VLMs to the THz domain and evaluate them under zero-shot and one-shot settings. Our results show that ICL improves classification and interpretability in low-data regimes. This is the first application of ICL-enhanced VLMs to THz imaging, offering a promising direction for resource-constrained scientific domains. Code: GitHub repository.

Abstract source ↗

Figures

Amplitude and phase images of a terahertz sample.
Figure 1. Example frame from the dataset (Frame Number 0663), selected from the total of 1,400 frames. This frame displays intensity (top) and phase (bottom) plots. Each plot shows a circular region on the left indicating the C4 explosive and a rectangular region on the right indicating the metal plate. While the actual experiment displays intensity and phase images side by side (intensity left, phase right), this figure arranges them vertically for better readability. View in source ↗
Example illustrating how crops were extracted from the 1,400 frames in the dataset.
Figure 2. Example illustrating how crops were extracted from the 1,400 frames in the dataset. On the left is the complete frame (Frame Number 0663, also shown in Figure 1) displaying intensity and phase plots with the visible “C4” circle. On the right is an enlarged view of a selected crop from this frame. View in source ↗

Cite this paper

Download .bib 
@inproceedings{poggi2025smarteyesfor70,
  title = {{Smart Eyes for Silent Threats - VLMs and In-Context Learning for THz Imaging}},
  author = {Nicolas Poggi and Shashank Agnihotri and Margret Keuper},
  booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVw 2025)},
  year = {2025},
  url = {https://arxiv.org/pdf/2507.15576?}
}

Figures and abstract are reproduced from the linked research sources. Credit remains with the authors and publishers.