I really enjoy reading and doing one 2d pose estimation project using PoolFormer as backbone, also love the idea of metaformer. Have you thought about pretraining the model using MAE? Would you expect to have a performance boost as ViT does? Thanks in advance
Thank you for your attention. I am so sorry for the late response. In the past several months, I didn't have large computation resources for this project to conduct the experiments.
Hi,
I really enjoy reading and doing one 2d pose estimation project using PoolFormer as backbone, also love the idea of metaformer. Have you thought about pretraining the model using MAE? Would you expect to have a performance boost as ViT does? Thanks in advance