Windaway

Results 10 comments of Windaway

You can regard the initialization is to make the feature of pixel to get information from its eight neighborhood.

不知道,这个东西训练太慢,我琢磨写个YOLOv3做个更快的图像增广,目前增广写完了,但是eager部分还待完成。

https://github.com/Windaway/FCRN-Depth-Prediction-Tensorflow I have writen a project for transfer learning.

https://github.com/Windaway/FCRN-Depth-Prediction-Tensorflow I have writen a project for transfer learning.

代码里写错,features和x2 x4 之类参数重复了

之前写CornerPool用了一个更脏的手法,直接边上填Batch最小,然后Maxpool。这样就没问题了。 https://tensorlayer.readthedocs.io/en/2.0.0/_modules/tensorlayer/layers/pooling.html#CornerPool2d

https://github.com/Windaway/DSOD-Tensorflow DSOD TF撒ver手without不any管support。

感谢,不过希望加入conda forge和pytorch等常用三方源

Remove normalize. Skimage will remove the features.