How saccadic vision might help with theinterpretability of deep networks

We describe how some problems (interpretability,lack of object-orientedness) of modern deep networks potentiallycould be solved by adapting a biologically plausible saccadicmechanism of perception. A sketch of such a saccadic visionmodel is proposed. Proof of concept experimental results areprovided to support the proposed approach.

cs.CV cs.AI Visual Attention and Saliency Detection Anomaly Detection Techniques and Applications Neural Networks and Applications

View graph Source

Paper

Similar papers

How saccadic vision might help with the interpretability of deep networksIana Sereda, Grigory V. Osipov2021 In Library

Saccade Mechanisms for Image Classification, Object Detection and TrackingSaurabh Farkya, Z. Daniels, Aswin Raghavan et al.2022 · arXiv.org · 3 citations In Library

Deep Learning Neural Networks as a Component of a Model of Saccadic GenerationSofia Krasovskaya, Georgii Zhulikov, W. Joseph MacInnes2021 · Research Papers in Economics · 0 citations In Library

From deep learning to mechanistic understanding in neuroscience: the structure of retinal predictionHidenori Tanaka, Aran Nayebi, Niru Maheswaranathan et al.2019 · Neural Information Processing Systems · 68 citations In Library

DISCOVER: Making Vision Networks Interpretable via Competition and DissectionKonstantinos P. Panousis, Sotirios Chatzis2023 · arXiv (Cornell University) · 9 citations In Library

A Generative Model of Cognitive State from Task and Eye MovementsW. Joseph MacInnes, A. Hunt, A. Clarke et al.2018 · Cognitive Computation · 15 citations In Library

Towards Interpretable and Attentive Deep Visual Perception Through Iterative Refinement in Convolutional Neural NetworksMohamed Nawas Paralath2025 In Library

Spatiotemporal evidence accumulation through saccadic sampling for object recognitionZhihao Zheng, Jia Hu, Gouki Okazawa2024 · bioRxiv (Cold Spring Harbor Laboratory) · 0 citations In Library

A dual foveal-peripheral visual processing model implements efficient saccade selectionEmmanuel Daucé, Pierre Albiges, Laurent Perrinet2020 · Journal of Vision In Library

On the visual analytic intelligence of neural networksStanisław Woźniak, Hlynur Jónsson, Giovanni Cherubini et al.2023 · Nature Communications In Library

Source of record

arXiv

44B

Welcome.

This panel sticks with you. Pick anything below and it opens right beside it, so you can dig through 60,000-plus records without ever losing your spot here.

Travel 44B

LibraryPapers, policy, standards, statute LabsEvery organization building AI ModelsIntelligence, price, and speed BenchmarksThe evaluation catalog IncidentsAI failures and harms SearchOne field across everything DashboardThe state of AI accountability in NY 44B RegistryThe Art. 44-B compliance portal