AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Optimizing Filter Size in Convolutional Neural Networks for Facial Action Unit Recognition

Shizhong Han, Zibo Meng, Zhiyuan Li, James O'Reilly, Jie Cai, Xiaofeng Wang, Yan Tong

CVPR 2018 · Architectures and efficiency › CNN design · approach: Compact CNN architectures · 69 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
learn filter sizes in all three convolutional layers instead of fixing themCNN on BP4D · BP4D developmentF1: 0.499 -> 0.515Improved
learn filter sizes in all three convolutional layers instead of fixing themCNN on BP4D · BP4D development2AFC: 0.664 -> 0.691Improved
learn the filter size only in the first convolutional layer rather than all three (ablation)OFS-CNN on BP4D · BP4D developmentF1: 0.515 -> 0.501Hurt
learn the filter size only in the first convolutional layer rather than all three (ablation)OFS-CNN on BP4D · BP4D development2AFC: 0.666 -> 0.691Hurt
optimize filter sizes rather than use fixed 5×5 filtersCNN on BP4D at 64×48 · BP4D development, 64×48F1: 0.499 -> 0.515Improved
optimize filter sizes rather than use fixed 5×5 filtersCNN on BP4D at 128×96 · BP4D development, 128×96F1: 0.522 -> 0.533Improved
optimize filter sizes rather than use fixed 5×5 filtersCNN on BP4D at 256×192 · BP4D development, 256×192F1: 0.478 -> 0.537Improved
replace fixed filter sizes with learned filter sizesCNN on DISFA · DISFAF1: 0.515 -> 0.553Improved
replace fixed filter sizes with learned filter sizesCNN on DISFA · DISFA2AFC: 0.850 -> 0.850No clear effect
use OFS-CNN instead of GoogLeNet's multiple-filter-size architectureAU recognition on BP4D · BP4DF1: 0.531 -> 0.537Improved
use OFS-CNN instead of GoogLeNetAU recognition on BP4D · BP4Dtesting speed: more than 6 times fasterImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy