search

icandle / CAMixerSR

CAMixerSR: Only Details Need More “Attention” (CVPR 2024)
https://arxiv.org/abs/2402.19289
Apache License 2.0
209 stars 11 forks source link

Flops #11

Closed guo794 closed 5 months ago

guo794 commented 5 months ago

Why is it that the model I got after changing [1,3,32,32] to [1,3,320,180] on the original code is much more computationally intensive than in the paper? computation

icandle commented 5 months ago

For Large-Image SR, the FLOPs are calculated as ClassSR. For lightweight SR, we use fvcore to directly compute the MAdds as previous lightweight models, which leads to the difference.

from fvcore.nn import FlopCountAnalysis, flop_count_table