SPANetV2 Official
August 10, 2026 ยท View on GitHub
๐ History:
- (2026.05.02) Accepted in IEEE TPAMI2026 ๐ฅณ
- (2025.03.31) arXiv2025: Spectral-Adaptive Modulation Networks for Visual Perception
- ICCV2023: SPANet: Frequency-balancing Token Mixer using Spectral Pooling Aggregation Modulation
๐ค It currently includes code and models for the following tasks:
Main results on ImageNet-1K
Please see image_classification for more details.
| Model | Pretrain | Resolution | Top-1 | #Params (M) | MACs (G) |
|---|---|---|---|---|---|
| SPANeV2-S18-pure | ImageNet-1K | 224x224 | 83.4 | 29 | 4.2 |
| SPANeV2-S18-hybrid | ImageNet-1K | 224x224 | 83.9 | 27 | 4.2 |
| SPANeV2-S36-pure | ImageNet-1K | 224x224 | 84.4 | 44 | 8.1 |
| SPANeV2-S36-hybrid | ImageNet-1K | 224x224 | 84.7 | 41 | 8.1 |
| SPANeV2-M36-pure | ImageNet-1K | 224x224 | 84.9 | 61 | 13.7 |
| SPANeV2-M36-hybrid | ImageNet-1K | 224x224 | 85.3 | 58 | 13.6 |
| SPANeV2-B36-pure | ImageNet-1K | 224x224 | 85.0 | 100 | 24.3 |
| SPANeV2-B36-hybrid | ImageNet-1K | 224x224 | 85.6 | 100 | 23.9 |
With the scaled sigmoid for SRF
| Model | Pretrain | Resolution | Top-1 | #Params (M) | MACs (G) |
|---|---|---|---|---|---|
| SPANeV2-S18-pure | ImageNet-1K | 224x224 | 83.5 | 29 | 4.2 |
| SPANeV2-S18-hybrid | ImageNet-1K | 224x224 | -.- | 27 | 4.2 |
| SPANeV2-S36-pure | ImageNet-1K | 224x224 | -.- | 44 | 8.1 |
| SPANeV2-S36-hybrid | ImageNet-1K | 224x224 | 84.9 | 41 | 8.1 |
Main results on COCO object detection and instance segmentation
Please see object_detection for more details.
Mask R-CNN 3x
| Backbone | box mAP | mask mAP | MACs (G) |
|---|---|---|---|
| SPANet-S18-pure | 48.0 | 42.9 | 255 |
| SPANet-S18-hybrid | 49.6 | 44.3 | 251 |
Cascade Mask R-CNN 3x
| Backbone | box mAP | mask mAP | MACs (G) |
|---|---|---|---|
| SPANet-S18-pure | 51.6 | 44.7 | 734 |
| SPANet-S18-hybrid | 52.8 | 45.7 | 729 |
| SPANet-S36-pure | 53.1 | 45.9 | 814 |
| SPANet-S36-hybrid | 53.8 | 46.4 | 806 |
Main results on ADE20K semantice segmentation
Please see semantic_segmentation for more details.
Semantic FPN
| Backbone | mIoU | #params (M) | MACs (G) |
|---|---|---|---|
| SPANet-S18-pure | 46.7 | 33 | 43 |
| SPANet-S18-hybrid | 47.8 | 31 | 48 |
| SPANet-S36-pure | 47.9 | 48 | 64 |
| SPANet-S36-hybrid | 48.6 | 45 | 74 |
UperNet
| Backbone | mIoU | #params (M) | MACs (G) |
|---|---|---|---|
| SPANet-S18-pure | 48.7 | 60 | 930 |
| SPANet-S18-hybrid | 49.1 | 58 | 925 |
| SPANet-S36-pure | 49.8 | 75 | 1012 |
| SPANet-S36-hybrid | 51.6 | 71 | 1004 |
โญ Cite SPANetV2
If you find this repository useful, please give us stars and use the following BibTeX entry for citation.
@article{yun2026spectral,
title={Spectral-Adaptive Modulation Networks for Visual Perception},
author={Yun, Guhnoo and Yoo, Juhan and Kim, Kijung and Lee, Jeongho and Seo, Paul Hongsuck and Kim, Dong Hwan},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2026},
publisher={IEEE}
}