Utilized a U-Net to perform segmentation of buildings in satellite imagery. Pre-trained the U-Net in a semi-supervised fashion by changing the last layer and training it as an autoencoder. Then switched last layer back to normal and trained on labelled data to perform the segmentation. Poster with results can be seen here.
-
Notifications
You must be signed in to change notification settings - Fork 5
conlamon/semi-supervised-satellite-image-segmentation
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
About
Exploration of using semi-supervised pre-training to improve convolutional neural network image segmentation performance on satellite imagery
Resources
Stars
Watchers
Forks
Releases
No releases published
Packages 0
No packages published