search

tist0bsc / SGCN

Split Depth-wise Separable Graph Convolution Network for Road Extraction in Complex Environment from High-resolution Remote Sensing Imagery
51 stars 11 forks source link

Prediction results #1

Open SonwYang opened 2 years ago

SonwYang commented 2 years ago

I train this model according to the default configuration. When the training is done, I run the prediction code and found that the the visualization of prediction is bad. Can you provide your best model params?

tist0bsc commented 2 years ago

I train this model according to the default configuration. When the training is done, I run the prediction code and found that the the visualization of prediction is bad. Can you provide your best model params?

Sorry, I upload the wrong dataset version, the validation set should be test set while the test set should be validation set. And I lost my previous model file, so I train a new one with 100 epochs. I put the best SGCN model (.pth), evalution results of new test set , visualization results of new test set, log file of new val set on https://pan.baidu.com/s/1ZiQcqEfdJa5HruxsUYobGQ , code:1234 and the dataset on Google Drive has been updated.

lyf6 commented 2 years ago

@SonwYang have you repeated this result?

yjwong1999 commented 1 month ago

Hi @tist0bsc

Can you share the pretrained model weights via google drive link? I'm unable to access your pretrained model via the baidu drive, because I couldn't register an account 🥲