Research map / Detection and segmentation / CNN features
Spectral-Adaptive Modulation Networks for Visual Perception
Guhnoo Yun, Juhan Yoo, Ki‐Jung Kim, Jeongho Lee, Paul Hongsuck Seo, Dong Hwan Kim
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 spectral rescaling filter (SRF) to SAG (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.2 -> 83.4 | Improved |
| use multi-scale kernels instead of single 3x3 kernel (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.1 -> 83.4 | Improved |
| use multi-scale kernels instead of single 5x5 kernel (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.1 -> 83.4 | Improved |
| use multi-scale kernels instead of single 9x9 kernel (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.2 -> 83.4 | Improved |
| use a single shared SRF instead of depth-wise SRF (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.4 -> 83.4 | No clear effect |
| replace SepConv with SPAM in all stages (ablation) | SPANetV2-S18 · ImageNet-1K | top-1 accuracy: 83.0 -> 83.4 | Improved |
| use SPAM in the first two stages instead of SepConv (ablation) | SPANetV2-S18 · ImageNet-1K | top-1 accuracy: 83.0 -> 83.1 | Improved |
| add Attention in the last two stages after SPAM stages (ablation) | SPANetV2-S18 · ImageNet-1K | top-1 accuracy: 83.4 -> 83.7 | Improved |
| use MixAttention in stage 3 and Attention in stage 4 (ablation) | SPANetV2-S18 · ImageNet-1K | top-1 accuracy: 83.4 -> 83.9 | Improved |
| use ResScale branch scaling instead of none (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.2 -> 83.4 | Improved |
| use ResScale instead of LayerScale or BranchScale (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.3 -> 83.4 | Improved |
| disable biases in Norm, FC and Conv (ablation) | SPANetV2-S18-pure · ImageNet-1K | top-1 accuracy: 83.3 -> 83.4 | Improved |
Closest papers by meaning
- SPANet: Frequency-balancing Token Mixer using Spectral Pooling Aggregation Modulation (ICCV 2023 · similarity 0.87)
- HoloNets: Spectral Convolutions do extend to Directed Graphs (ICLR 2024 · similarity 0.86)
- Self-Attention-Based Contextual Modulation Improves Neural System Identification (ICLR 2025 · similarity 0.85)
- SONIC: Spectral Oriented Neural Invariant Convolutions (ICLR 2026 · similarity 0.85)
- Spectral Prompt Tuning: Unveiling Unseen Classes for Zero-Shot Semantic Segmentation (AAAI 2024 · similarity 0.84)
- Any-Resolution AI-Generated Image Detection by Spectral Learning (CVPR 2025 · similarity 0.83)
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.
