Aerial-Imagery-Segmentation

Aerial imagery of Dubai obtained by MBRSC satellites (kaggle Dataset) annotated with pixel-wise semantic segmentation in 6 classes.

1- Building: #3C1098

2- Land (unpaved area): #8429F6

3- Road: #6EC1E4

4- Vegetation: #FEDD3A

5- Water: #E2A929

6- Unlabeled: #9B9B9B

Random Images and its masks from Dataset

Predictions

the colors of pred mask is different from true mask bacause i have assigned each class to a specfic integer from 0 to 5 (6 classes) and didn't convert it back to its RGB values

Model Accuracy

Model Loss

Model Performance per Epoch (usng CSVLogger Callback)

train_performance_per_epoch.csv