search

xuebinqin / DIS

This is the repo for our new project Highly Accurate Dichotomous Image Segmentation
Apache License 2.0
2.28k stars 270 forks source link

GT_Encoder validated against training dataset #45

Open Ozwol opened 2 years ago

Ozwol commented 2 years ago

Hi, looking at train_valid_inference_main.py (lines 131-132), GT_Encoder training is validated against training dataset instead of validation dataset (that has been commented out). Why have you make this choice? Thanks

xuebinqin commented 2 years ago

Because our gt encoder is to "represent" the ground truth masks in high-dimensional space by "overfitting" them. That validation code is copied from the segmentation function, so we commented it out and validate the gt encoder on training set.

On Wed, Sep 14, 2022 at 7:53 AM Ozwol @.***> wrote:

Hi, looking at train_valid_inference_main.py https://github.com/xuebinqin/DIS/blob/main/IS-Net/train_valid_inference_main.py#L131 (lines 131-132), GT_Encoder training is validated against training dataset instead of validation dataset (that has been commented out). Why have you make this choice? Thanks

— Reply to this email directly, view it on GitHub https://github.com/xuebinqin/DIS/issues/45, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSGORNHZ7PT2CF3VYDPDC3V6HRHDANCNFSM6AAAAAAQMP5K2Y . You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- Xuebin Qin PhD Department of Computing Science University of Alberta, Edmonton, AB, Canada Homepage: https://xuebinqin.github.io/