search

TreB1eN / InsightFace_Pytorch

Pytorch0.4.1 codes for InsightFace
MIT License
1.72k stars 418 forks source link

Fine tuning model by my own custom dataset. #175

Open sangtv9 opened 2 years ago

sangtv9 commented 2 years ago

Thanks for you great repo. I used pretrained model for my dataset. But, It does not reach good enough accuracy. So I must re-train the model by my dataset. But I only reach 65% at accuracy on validation dataset. The model loss only decrease to approximate 10. Have anyone met the issue? Have you had any idea to deal this? Thanks very much.

sangtv9 commented 2 years ago

This is my config for training SGD ( Parameter Group 0 dampening: 0 lr: 0.05 momentum: 0.9 nesterov: False weight_decay: 0.0004 Parameter Group 1 dampening: 0 lr: 0.05 momentum: 0.9 nesterov: False weight_decay: 0 )
I use CrossEntropy for loss function

momohuangsha commented 2 years ago

@sangtv9 Hey! Has your problem been solved? whta is your backbone network? I have the same problem (backbone:mobileface)

ANDRESHZ commented 11 months ago

Hi @momohuangsha and @sangtv9, I hope you are having nice results, I am wondering if you finally found a way to tune with your dataset.

Can you share with the community your code?