Multiclass Segmentation using UNet for autonomous cars

1. The notebook unet-keras contains the code for the segmenation model build in keras deep learning library.
2. The data used for the multiclass segmentation was obtained from the Cityscapes dataset. The dataset can be found here.
3. The training was performed on Google Colab.
4. The original dataset contains 30 classes, which were cut down to 8 for the purpose of training this model, due to lack of efficient resources for training.
5. The selected classes for training include:
   • road
   • sidewalk
   • person
   • car
   • truck
   • bus
   • sky
   • ground