search

HS-YN / PAVER

Official repository of Panoramic Vision Transformer for Saliency Detection in 360° Videos (ECCV 2022)
MIT License
33 stars 5 forks source link

How to reproduce the performance in the paper？ #3

Closed liuhuichuan closed 1 year ago

liuhuichuan commented 1 year ago

I downloaded the model, and run train,py and my own code. However, CC and other metrics are all smaller than them in the paper. Do i need to train it?

liuhuichuan commented 1 year ago

and i have another question, why is lr 2e-7? I use lr as 1e-3, but in the first 13 epoch, the CC change little, almost the same value. Can you tell me some explanation? thanks

liuhuichuan commented 1 year ago

after i tried sometimes， i got the correct results. Maybe need tries more times?