problems about training

layumi / Seg-Uncertainty

IJCAI2020 & IJCV2021 :city_sunrise: Unsupervised Scene Adaptation with Memory Regularization in vivo

https://arxiv.org/abs/1912.11164

MIT License

386 stars 51 forks source link

problems about training #11

Closed Serge-weihao closed 4 years ago

Serge-weihao commented 4 years ago

why you train your model in stage 1 and 3 with 100000 iterations, but choose 25000th iteration for finetunning and final evaluation?

layumi commented 4 years ago

@Serge-weihao Early-stopping is to prevent the model overfitting to the source dataset, which is widely-used in previous works. You may check the https://github.com/RoyalVane/CLAN

Serge-weihao commented 4 years ago

does it mean the process after 25000th is useless?

layumi commented 4 years ago

@Serge-weihao Yes. You may skip the training.