arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Author

Timothy Player

3 papers · 2022–2023

2022: 22023: 1
2022peak 2 in 20222023

Frequent co-authors

Papers

showing 1–3 of 3 RSS

Dataset Augmentation in Papyrology with Generative Models: A Study of Synthetic Ancient Greek Character Images

Matthew I. Swindall, Timothy Player, Ben Keener +7 more

Computer Vision and Pattern Recognition · Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence · · cited 7× · DOI ↗ · more like this ·

Character recognition models rely substantially on image datasets that maintain a balance of class samples. However, achieving a balance of classes is particularly challenging for ancient manuscript contexts as character instances may be significantly limited. In this paper, we present findings from a study that assess the efficacy of using synthetically generated character instances to augment an existing dataset of ancient Greek character images for use in machine learning models. We complement our model exploration by engaging professional papyrologists to better understand the practical opportunities afforded by synthetic instances. Our results suggest that synthetic instances improve model performance for limited character classes, and may have unexplored effects on character classes more generally. We also find that trained papyrologists are unable to distinguish between synthetic…

Incorporating Crowdsourced Annotator Distributions into Ensemble Modeling to Improve Classification Trustworthiness for Ancient Greek Papyri

Graham West, Matthew I. Swindall, Ben Keener +4 more

cs.CV · · cited 0× · arXiv ↗ · more like this ·

Performing classification on noisy, crowdsourced image datasets can prove challenging even for the best neural networks. Two issues which complicate the problem on such datasets are class imbalance and ground-truth uncertainty in labeling. The AL-ALL and AL-PUB datasets - consisting of tightly cropped, individual characters from images of ancient Greek papyri - are strongly affected by both issues. The application of ensemble modeling to such datasets can help identify images where the ground-truth is questionable and quantify the trustworthiness of those samples. As such, we apply stacked generalization consisting of nearly identical ResNets with different loss functions: one utilizing sparse cross-entropy (CXE) and the other Kullback-Liebler Divergence (KLD). Both networks use labels drawn from a crowd-sourced consensus. This consensus is derived from a Normalized Distribution of…

From Concept to Field Tests: Accelerated Development of Multi-AUV Missions Using a High-Fidelity Faster-than-Real-Time Simulator

Timothy R. Player, Arjo Chakravarty, Mabel M. Zhang +4 more

cs.RO · · arXiv ↗ · more like this ·

We designed and validated a novel simulator for efficient development of multi-robot marine missions. To accelerate development of cooperative behaviors, the simulator models the robots' operating conditions with moderately high fidelity and runs significantly faster than real time, including acoustic communications, dynamic environmental data, and high-resolution bathymetry in large worlds. The simulator's ability to exceed a real-time factor (RTF) of 100 has been stress-tested with a robust continuous integration suite and was used to develop a multi-robot field experiment.