arXiv subjects

Samuel Pantze

Publications and source records attributed to Samuel Pantze.

3 recordsLinked to original sources

arXiv2026

Metric-Guided Synthetic Image Data Rendering for Deep Learning compatible with Agentic AI

Deep learning computer vision for scientific applications requires collecting and annotating large datasets in a laborious, expensive and error-prone process. Synthetic data generation through 3D modelling and rendering may simplify this process and increase the accuracy of annotations by generating them programmatically. However, minimising the domain gap between real and synthetic images visually is subjective and lacks systematic quantitative guidance. We present GraNatPy, a Python package with metrics to guide improvement of the rendered scene. We show that quantifiable increase in realism, diversity and size of rendered dataset correlates with improved visual perception of the scene and higher zero-shot performance of an object detection model. Furthermore, we demonstrated using photographs of virological plaque assays that gradient similarity affects performance on small object detection, which can be improved by mixing real and synthetic data. Finally, we turn procedural data rendering into an agentic skill (SynthClaw) to automate the procedural parameter optimisation.

cs.CV↗

arXiv2026

SciFi-VIS: Way Out There -- How SciFi and Visualization Influence Each Other

We propose a hybrid half-day workshop at IEEE VIS 2026, calling for participation from visualization researchers and science fiction creators in order to develop a systematic understanding of the two-way relationship these communities have long shared. We invite submissions of creative formats showcasing connections and inspiring future research. Our workshop plan includes a keynote, lightning talks, brainstorming, cross-community critique, affinity mapping, and discussion around identified themes.

cs.HC↗

arXiv2025

manvr3d: A Platform for Human-in-the-loop Cell Tracking in Virtual Reality

We propose manvr3d, a novel VR-ready platform for interactive human-in-the-loop cell tracking. We utilize VR controllers and eye-tracking hardware to facilitate rapid ground truth generation and proofreading for deep learning-based cell tracking models. Life scientists reconstruct the developmental history of organisms on the cellular level by analyzing 3D time-lapse microscopy images acquired at high spatio-temporal resolution. The reconstruction of such cell lineage trees traditionally involves tracking individual cells through all recorded time points, manually annotating their positions, and then linking them over time to create complete trajectories. Deep learning-based algorithms accelerate this process, yet depend heavily on manually-annotated high-quality ground truth data and curation. Visual representation of the image data in this process still relies primarily on 2D renderings, which greatly limits spatial understanding and navigation. In this work, we bridge the gap between deep learning-based cell tracking software and 3D/VR visualization to create a human-in-the-loop cell tracking system. We lift the incremental annotation, training and proofreading loop of the deep learning model into the 3rd dimension and apply natural user interfaces like hand gestures and eye tracking to accelerate the cell tracking workflow for life scientists.

cs.HC↗