Architecture & Feature Aggregation
Image Super-Resolution with SMFANet
163
•
280
Apache-2.0
Image-to-Image
PyTorch
English
Opened about 10 months ago
by @messe-7257
2
@messe-725710 months ago
If you had to pick one feature that makes SMFANet’s SMFA block superior to other multi-scale fusion modules, what would it be?
Hi @messe-7257,
The SMFA block dynamically re-weights and fuses features across scales by learning channel-wise modulation masks. Unlike static concatenation or addition, SMFA tailors the contribution of each feature map at runtime, enhancing context adaptivity with minimal extra parameters.
You can refer to this information!