Semantic-Segmentation

Introduction

The goal is to perfrom semantic segmentation on the Oxford-IIIT Pet Dataset.

Model

U-Net based Archtircure impemeted from scratch using pytorch-lightning found in Unet.py.

RESULTS

The model achived a mean IoU of 0.8495 and a mean Dice Coefficient of 0.8861 on the validation set.

Training

To train the model, run train.py with the following arguments:

python train.py --log_dir /path/to/log/dir