Understanding the Performance of Deep Computer Vision Models: A Symbolic Regression Approach to Accuracy and Latency Prediction
Sensors · 27.06.2026 · 26 (13) · 4093
DOI: 10.3390/s26134093 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 27.06.2026. Verfasst von Divyesh Rameshbhai Dhanani und 5 weiteren Autor:innen. Beteiligte Einrichtung: Computer Vision Lab, Center for Artificial Intelligence and Data Science (CAIDAS) and Institute of Computer Science (IFI), University of Würzburg, Jo…. Die Arbeit nennt 57 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Alexander von Humboldt Foundation.
Abstract (Originalsprache)
Deploying deep vision models on edge hardware requires understanding how architecture and training hyperparameters jointly determine accuracy and inference latency, yet these relationships remain poorly characterized in a systematic, data-driven manner. This paper presents a two-stage statistical framework providing interpretable, closed-form insights into both. In the first stage, we apply distance correlation (dCor) and the maximal information coefficient (MIC) across seven image-classification datasets, revealing that batch size and total layer count are the strongest universal accuracy predictors (mean dCor: 0.228 and 0.174), while learning rate achieves the highest MIC (0.226), reflecting a non-monotonic relationship with accuracy. In the second stage, PySR symbolic regression (representing, to our knowledge, the first application to cross-dataset vision model accuracy prediction) derives compact, interpretable formulas. Dataset-specific models achieve R2 from 0.20 to 0.45; a universal model achieves a mean leave-one-dataset-out R2 of 0.23, remaining strictly positive on all held-out datasets, whereas ordinary linear regression collapses to R2=−0.71. We further derive device-specific inference latency formulas for CPU, GPU, and NPU, outperforming classical baselines by 6.7×–14.8× in R2 and confirming fundamental device heterogeneity. Together, these results offer interpretable surrogate models for screening deep vision architectures under accuracy and latency constraints in edge deployment.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Computer Vision Lab, Center for Artificial Intelligence and Data Science (CAIDAS) and Institute of Computer Science (IFI), University of Würzburg, John-Skilton-Straße 4a, 97074 Würzburg, Germany.
ORCID-Profile: Divyesh Rameshbhai Dhanani, Faraz Kayani, Saif U Din, Alice Arslanian, Dmitry Ignatov, Radu Timofte
Förderung
- Alexander von Humboldt Foundation
Zitieren
APA
Dhanani, D. R., Kayani, F., Din, S. U., Arslanian, A., Ignatov, D., & Timofte, R. (2026). Understanding the Performance of Deep Computer Vision Models: A Symbolic Regression Approach to Accuracy and Latency Prediction. Sensors, 26(13), 4093. https://doi.org/10.3390/s26134093
Vancouver
Dhanani DR, Kayani F, Din SU, Arslanian A, Ignatov D, Timofte R. Understanding the Performance of Deep Computer Vision Models: A Symbolic Regression Approach to Accuracy and Latency Prediction. Sensors. 2026;26(13):4093. doi:10.3390/s26134093
BibTeX
@article{Dhanani2026,
author = {Dhanani, Divyesh Rameshbhai and Kayani, Faraz and Din, Saif U and Arslanian, Alice and Ignatov, Dmitry and Timofte, Radu},
title = {Understanding the Performance of Deep Computer Vision Models: A Symbolic Regression Approach to Accuracy and Latency Prediction},
journal = {Sensors},
year = {2026},
volume = {26},
number = {13},
pages = {4093},
doi = {10.3390/s26134093}
}
EUVIMED