- Abstract
U-NET model for RGB images, using aerial images as an example. - Codes
1. unet_model
Code for UNET model, includes data enhancement(data.py), model(model.py), train(main.py) and prediction(prediction.py).
2. post_process(Optional)
Code for post process, includes smooth(smooth.py) and binary(binary.py), run main.py to use them. - Data
Example data -- https://project.inria.fr/aerialimagelabeling/
Also providing a short code (/data/segmentation256.py) to cut TIFF images to size 256*256. - Result
Acc. -- 94% IOU -- 91
RilaShu/unet-model-aerial-images
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|