Semantic segmentation of remote sensing images.

1. Operation system: Ubuntu 16.04.3 LTS
   
2. CUDA version: 8.0
   
3. Programming language: Python 2.7
   Modules:
   tensorflow-gpu==1.4.1, scikit-image==0.13.0, tifffile==0.13.5,
   pandas==0.20.3, keras==1.2.2, h5py==2.7.0, tqdm== 4.19.4
   

• Reference: kaggle-dstl 3rd blog
  
• Framework:
  
  
• Input and output: 1024*1024*3 (in), 1000*1000*1 (out), row*column*channel

left: image, middle: ground truth(label), right: prediction

1. Airport
   
   
2. Baresoil
   
   
3. Building
   
   
4. Farmland
   
   
5. Road
   
   
6. Vegetation
   
   
7. Water