Iris Segmentation Using Segnet

This repo borrows codes from https://github.com/meetshah1995/pytorch-semseg and https://github.com/utkuozbulak/pytorch-custom-dataset-examples.

Given that Iris dataset has been downloaded as "Iris_Segmentation_Dataset.tar.gz".

Requirement

Pytorch 0.4.1

Pandas 0.20.3

Pillow 5.2.0

How to Run

1. Make a new dir named "data"

   # From segnet_iris/
   mkdir data

2. Extract the dataset into data folder

   # From segnet_iris/data/
   tar -zvxf Iris_Segmentation_Dataset.tar.gz

3. Copy eval.csv and train.csv from segnet_iris to segnet_iris/data/iris_segmentation/lists/

   (you may also run divide_iris_data.py from src to generate new lists)

4. For training, run train.py. You may modify data loader's worker number(line 19) according to your CPU cores and CUDA device number(line 23) according to your GPU number.

   # From segnet_iris/src/
   python train.py

5. For evaluation, modify eval.py according to the best model you have got. Then make a dir named output in segnet_iris. Run eval.py. As previous said, you shold adjust data loader's worker number and CUDA device number according to your machine.

   # From segnet_iris/src/
   python eval.py