Research map / Detection and segmentation / CNN features
Delving Deeper into Anti-Aliasing in ConvNets
Xueyan Zou, Fanyi Xiao, Zhiding Yu, Yuheng Li, Yong Jae Lee
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add predicted content-aware filters in place of fixed LPF (3x3) | ResNet-101 · ImageNet val | top-1 accuracy: 78.4 -> 79.0 | Improved |
| add adaptive anti-aliasing module to ResNet-101 (3x3) | ResNet-101 · ImageNet val | top-1 accuracy: 77.7 -> 79.0 | Improved |
| use 5x5 instead of 3x3 learned filter (ablation) | ResNet-101 · ImageNet val | top-1 accuracy: 79.0 -> 78.6 | Hurt |
| use 5x5 fixed LPF instead of 3x3 (ablation) | ResNet-101 · ImageNet val | top-1 accuracy: 78.4 -> 77.7 | Hurt |
| add adaptive anti-aliasing module (3x3) | ResNet-101 · ImageNet to ImageNet VID | top-1 accuracy: 67.6 -> 69.9 | Improved |
| add adaptive anti-aliasing instead of LPF | ResNet-101 · ImageNet to ImageNet VID | top-1 accuracy: 68.8 -> 69.9 | Improved |
| add adaptive anti-aliasing module | Mask R-CNN · MS COCO instance segmentation | mask mAP: 36.1 -> 37.2 | Improved |
| add adaptive anti-aliasing module | Mask R-CNN · MS COCO instance segmentation | mask mAISC: 62.9 -> 67.0 | Improved |
| replace LPF with adaptive anti-aliasing | Mask R-CNN · MS COCO instance segmentation | mask mAP: 36.8 -> 37.2 | Improved |
| add adaptive anti-aliasing module | Deeplab v3+ · PASCAL VOC 2012 | mIoU: 78.5 -> 80.3 | Improved |
| add adaptive anti-aliasing module | Deeplab v3+ · Cityscapes | mIoU: 78.5 -> 79.5 | Improved |
| add adaptive anti-aliasing module | Deeplab v3+ · Cityscapes | mASSC: 96.0 -> 96.3 | Improved |
Closest papers by meaning
- Making Convolutional Networks Shift-Invariant Again (ICML 2019 · similarity 0.89)
- Revisiting Downsampling in Semantic Segmentation: Fighting Aliasing with Dynamic Gaussian and Gabor Frequency Filters (AAAI 2026 · similarity 0.88)
- When Semantic Segmentation Meets Frequency Aliasing (ICLR 2024 · similarity 0.86)
- Group Downsampling with Equivariant Anti-aliasing (ICLR 2025 · similarity 0.85)
- Alias-Free ViT: Fractional Shift Invariance via Linear Attention (NeurIPS 2025 · similarity 0.85)
- Deep Edge Filter: Return of the Human-Crafted Layer in Deep Learning (NeurIPS 2025 · similarity 0.85)
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.
