Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866909958882197504 |
|---|---|
| author | Jérémie, Jean-Nicolas Daucé, Emmanuel Perrinet, Laurent U |
| author_facet | Jérémie, Jean-Nicolas Daucé, Emmanuel Perrinet, Laurent U |
| contents | From falcons spotting preys to humans recognizing faces, rapid visual abilities depend on a foveated retinal organization which delivers high-acuity central vision while preserving low-resolution periphery. This organization is conserved along early visual pathways but remains underexplored in machine learning. Here we examine how embedding a foveated retinotopic transformation as a preprocessing layer impacts convolutional neural networks (CNNs) for image classification. By applying a log-polar mapping to off-the-shelf models and retraining them, we retain comparable accuracy while improving robustness to scale and rotation. We show that this architecture becomes highly sensitive to fixation-point shifts, and that this sensitivity yields a proxy for defining saliency maps that effectively facilitates object localization. Our results show that foveated retinotopy encodes prior geometric knowledge, offering a solution to visual-search and enhancing both classification and localization. These findings connect biological vision principles with artificial networks, pointing to new, robust and efficient directions for computer-vision systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_15480 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks Jérémie, Jean-Nicolas Daucé, Emmanuel Perrinet, Laurent U Computer Vision and Pattern Recognition Neurons and Cognition From falcons spotting preys to humans recognizing faces, rapid visual abilities depend on a foveated retinal organization which delivers high-acuity central vision while preserving low-resolution periphery. This organization is conserved along early visual pathways but remains underexplored in machine learning. Here we examine how embedding a foveated retinotopic transformation as a preprocessing layer impacts convolutional neural networks (CNNs) for image classification. By applying a log-polar mapping to off-the-shelf models and retraining them, we retain comparable accuracy while improving robustness to scale and rotation. We show that this architecture becomes highly sensitive to fixation-point shifts, and that this sensitivity yields a proxy for defining saliency maps that effectively facilitates object localization. Our results show that foveated retinotopy encodes prior geometric knowledge, offering a solution to visual-search and enhancing both classification and localization. These findings connect biological vision principles with artificial networks, pointing to new, robust and efficient directions for computer-vision systems. |
| title | Foveated Retinotopy Improves Classification and Localization in Convolutional Neural Networks |
| topic | Computer Vision and Pattern Recognition Neurons and Cognition |
| url | https://arxiv.org/abs/2402.15480 |