search

charlesCXK / TorchSemiSeg

[CVPR 2021] CPS: Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision
MIT License
504 stars 74 forks source link

Memory and Inf time (fps) for CPS #81

Closed JoyHuYY1412 closed 1 year ago

JoyHuYY1412 commented 1 year ago

Hi Xiaokang,

Thank you for your great work! Would you mind providing me with the memory and inf time for CPS? It would help a lot for citing your work.
I appreciate your help.

JoyHuYY1412 commented 1 year ago

I think the memory is not just the number of trainable parameters as you used cutmix, which actually doubled the batch_size when feeding them into the network.

charlesCXK commented 1 year ago

Hi, we didn't count these numbers. But I agree with what you said, memory should be the situation of GPU memory usage during training, not the number of parameters.