search

IIGROUP / MANIQA

[CVPRW 2022] MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment
Apache License 2.0
307 stars 36 forks source link

Question for train the model on koniq #42

Closed cyclopssssss closed 1 year ago

cyclopssssss commented 1 year ago

Hi! use your source code of koniq checkpoint to train, but the plcc/srcc result is lower than yours 7ZQ4G{U9G4XAAC4)PK {QTJ how to train the model?

TianheWu commented 1 year ago

resize the input image to 224x224 and train

TianheWu commented 1 year ago

not crop

cyclopssssss commented 1 year ago

not crop

thanks so much for your reply~

cyclopssssss commented 1 year ago

but in the code of koniq dataset, resize is already done B8FX5V1V~T)QH42MDB732UK and in that case the crop will do nothing with the img JGTRE` PJ2OPH{1S3` ZM`X Any other advice about training? Thanks again for your patience!

TianheWu commented 1 year ago

Thanks! This result is within the acceptable fluctuation range. You can try to change the split seed to see if there are other results.

cyclopssssss commented 1 year ago

Got it and sincere thank for you!