My take on the FloodNet Challange 2021
CONTENTS
-
Data Resizing and Augmentation
- In this part i first resized the dataset and saved to my local devices or drive
- Then Augmentation and Transforming the data with one-hot encoding
- Implemented under-sampling, over-sampling and class weights for class inbalance problem
-
Model
- In this part I implemented 2 different models firstly my own U-net model and secondly PSPNet model implemented using Segmentation models for Tensorflow
- Also the dataset augmentation is called and set in this part
-
Training the models
-
Visualization of Predictions with plots
At total I have implemented 2 models (U-net/PSPNet) and one-hot encoding for dataset combined with under-sampling and over-sampling for class inbalance problems.