Research map / Architectures and efficiency / Attention and transformers
Adaptive Frequency Filters As Efficient Global Token Mixers
Zhipeng Huang, Zhizheng Zhang, Cuiling Lan, Zheng-Jun Zha, Yan Lu, Baining Guo
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 AFF token mixers to a baseline with no spatial token mixer (ablation) | AFFNet baseline · ImageNet-1K | top-1 accuracy: 77.9 -> 79.8 | Improved |
| use AFF token mixer instead of 3x3 conv token mixer (ablation) | AFFNet · ImageNet-1K | top-1 accuracy: 78.6 -> 79.8 | Improved |
| apply adaptive filtering in the frequency domain instead of the original domain (ablation) | AFF token mixer · ImageNet-1K | top-1 accuracy: 78.4 -> 79.8 | Improved |
| use Hadamard product instead of summation in frequency filtering (ablation) | AFF token mixer · ImageNet-1K | top-1 accuracy: 78.8 -> 79.8 | Improved |
| use AFF design instead of AFNO frequency operation (ablation) | AFFNet · ImageNet-1K | top-1 accuracy: 78.8 -> 79.8 | Improved |
| use AFF design instead of FNO (trainable matrices) (ablation) | AFFNet · ImageNet-1K | top-1 accuracy: 79.7 -> 79.8, params 141.0M -> 5.5M | Improved |
| use channel-specific masks instead of channel-shared masks (ablation) | AFF token mixer · ImageNet-1K | top-1 accuracy: 79.3 -> 79.8 | Improved |
| increase spatial kernel size in mask network to 3x3 (ablation) | AFF mask network · ImageNet-1K | top-1 accuracy: 79.8 -> 79.8, params 5.5M -> 7.9M | No clear effect |
| use regular linear layers (N_group=1) in mask network (ablation) | AFF mask network · ImageNet-1K | top-1 accuracy: 79.8 -> 79.9, params 5.5M -> 7.7M | No clear effect |
| use depthwise setting N_group=C in mask network (ablation) | AFF mask network · ImageNet-1K | top-1 accuracy: 79.8 -> 79.4 | Hurt |
| use AFFNet backbone instead of EdgeNext in SSD | EdgeNext · COCO detection | mAP: 27.9 -> 28.4 | Improved |
| use AFFNet backbone instead of EdgeNext in DeepLabv3 | EdgeNext · PASCAL VOC | mIoU: 80.2 -> 80.5 | Improved |
Closest papers by meaning
- Multiscale Low-Frequency Memory Network for Improved Feature Extraction in Convolutional Neural Networks (AAAI 2024 · similarity 0.87)
- Frequency-Adaptive Dilated Convolution for Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Frequency Dynamic Convolution for Dense Image Prediction (CVPR 2025 · similarity 0.86)
- FFT-Based Dynamic Token Mixer for Vision (AAAI 2024 · similarity 0.85)
- Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning (AAAI 2024 · similarity 0.85)
- Frequency-Aware Token Reduction for Efficient Vision Transformer (NeurIPS 2025 · similarity 0.84)
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.
