EagleConv: Bio-inspired Dual-Foveated Convolution for Robust Small Object Detection

Jianwei Liu, Lifei Hao, Baoqi Huang, Bing Jia, Xuandong Zhao
Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, PMLR 337:3922-3932, 2026.

Abstract

Small object detection underpins wide-area vision tasks such as UAV and remote-sensing imagery. However, standard convolutions often exhibit low-pass smoothing behavior, which suppresses the sparse edge cues of tiny targets and may cause them to be overwhelmed by background clutter, leading to semantic information loss at the early stages of feature extraction. Inspired by the dual-fovea physiology of the eagle eye, we propose EagleConv, a bio-inspired operator that superimposes three {Gaussian} components into a center-excitation, surround-inhibition, peripheral-context response profile to selectively amplify small-object signals while attenuating noise. Architecturally, EagleConv adopts a sparse dual-pathway design that integrates partial-channel depthwise dual-foveated filtering, pointwise channel mixing, and adaptive residual gating, followed by anti-aliasing downsampling for robust dimensionality reduction. Experiments on three public benchmarks show that augmenting YOLOv11s with EagleConv delivers consistent and significant gains over multiple convolution-augmentation baselines, achieving state-of-the-art performance and confirming its transferability and generality as a plug-and-play module for wide-area small-object detection.

Cite this Paper

BibTeX
@InProceedings{pmlr-v337-liu26a, title = {EagleConv: Bio-inspired Dual-Foveated Convolution for Robust Small Object Detection}, author = {Liu, Jianwei and Hao, Lifei and Huang, Baoqi and Jia, Bing and Zhao, Xuandong}, booktitle = {Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence}, pages = {3922--3932}, year = {2026}, editor = {Perković, Emilija and Malinsky, Daniel}, volume = {337}, series = {Proceedings of Machine Learning Research}, month = {17--21 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v337/main/assets/liu26a/liu26a.pdf}, url = {https://proceedings.mlr.press/v337/liu26a.html}, abstract = {Small object detection underpins wide-area vision tasks such as UAV and remote-sensing imagery. However, standard convolutions often exhibit low-pass smoothing behavior, which suppresses the sparse edge cues of tiny targets and may cause them to be overwhelmed by background clutter, leading to semantic information loss at the early stages of feature extraction. Inspired by the dual-fovea physiology of the eagle eye, we propose EagleConv, a bio-inspired operator that superimposes three {Gaussian} components into a center-excitation, surround-inhibition, peripheral-context response profile to selectively amplify small-object signals while attenuating noise. Architecturally, EagleConv adopts a sparse dual-pathway design that integrates partial-channel depthwise dual-foveated filtering, pointwise channel mixing, and adaptive residual gating, followed by anti-aliasing downsampling for robust dimensionality reduction. Experiments on three public benchmarks show that augmenting YOLOv11s with EagleConv delivers consistent and significant gains over multiple convolution-augmentation baselines, achieving state-of-the-art performance and confirming its transferability and generality as a plug-and-play module for wide-area small-object detection.} }
Endnote
%0 Conference Paper %T EagleConv: Bio-inspired Dual-Foveated Convolution for Robust Small Object Detection %A Jianwei Liu %A Lifei Hao %A Baoqi Huang %A Bing Jia %A Xuandong Zhao %B Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2026 %E Emilija Perković %E Daniel Malinsky %F pmlr-v337-liu26a %I PMLR %P 3922--3932 %U https://proceedings.mlr.press/v337/liu26a.html %V 337 %X Small object detection underpins wide-area vision tasks such as UAV and remote-sensing imagery. However, standard convolutions often exhibit low-pass smoothing behavior, which suppresses the sparse edge cues of tiny targets and may cause them to be overwhelmed by background clutter, leading to semantic information loss at the early stages of feature extraction. Inspired by the dual-fovea physiology of the eagle eye, we propose EagleConv, a bio-inspired operator that superimposes three {Gaussian} components into a center-excitation, surround-inhibition, peripheral-context response profile to selectively amplify small-object signals while attenuating noise. Architecturally, EagleConv adopts a sparse dual-pathway design that integrates partial-channel depthwise dual-foveated filtering, pointwise channel mixing, and adaptive residual gating, followed by anti-aliasing downsampling for robust dimensionality reduction. Experiments on three public benchmarks show that augmenting YOLOv11s with EagleConv delivers consistent and significant gains over multiple convolution-augmentation baselines, achieving state-of-the-art performance and confirming its transferability and generality as a plug-and-play module for wide-area small-object detection.
APA
Liu, J., Hao, L., Huang, B., Jia, B. & Zhao, X.. (2026). EagleConv: Bio-inspired Dual-Foveated Convolution for Robust Small Object Detection. Proceedings of the 42nd Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 337:3922-3932 Available from https://proceedings.mlr.press/v337/liu26a.html.

Related Material