What is it about?
Computer Vision AI suffers when the input images are noisy, as normally happens in the real world. In this paper we propose to clean them before further processing, increasing the accuracy at minimal cost. The core idea is to consider an explicit parametric representation of the noise, rather than relying on black-box architectures.
Featured Image
Photo by Jatniel Tunon on Unsplash
Why is it important?
Vision-based AI nowadays is present in almost all personal devices, making the task of improving its reliability a very desirable objective. Common approaches for this task have a series of downsides, a common one the computational cost. In our work we introduce a very small and efficient alternative.
Perspectives
Writing this article was a great pleasure as it has co-authors with whom I have had long standing collaborations and the results obtained were very promising. I hope it can inspire other people to start their journey into AI research.
Francesco Barbato
Universita degli Studi di Padova
Read the Original
This page is a summary of: A Modular System for Enhanced Robustness of Multimedia Understanding Networks via Deep Parametric Estimation, April 2024, ACM (Association for Computing Machinery),
DOI: 10.1145/3625468.3647623.
You can read the full text:
Resources
Contributors
The following have contributed to this page
