__init__.py

import torch
import torchvision
from torchvision import transforms
from torch import Tensor

import pandas as pd
from skimage import io
import matplotlib.pyplot as plt
from pathlib import Path

# Ignore warnings
import warnings
warnings.filterwarnings("ignore")

plt.ion()   # interactive mode
    
class MORRIS(torch.utils.data.Dataset):
    _storage_csv='morris.csv'
    _storage_jpg='jpgs'
    def __init__(self,root=Path(__file__).parent,transform=False):
        self.storage = Path(root)
        self.transform = transform
        self.index = pd.read_csv(self.storage / self._storage_csv)
    
    def __len__(self):
        return len(self.index)
    
    def __getitem__(self,idx):
        item = self.index.iloc[idx].to_dict()
        image = io.imread(self.storage / self._storage_jpg / self.index.iloc[idx].filename)
    
        if self.transform:
            image = self.transform(image)
        
        item['image'] = image
        return item
    
    def showitem(self,idx):
        plt.imshow(transforms.ToPILImage()(self.__getitem__(idx)['image']))
    
    def _self_validate(self):
        """try loading each image in the dataset"""
        allgood=True
        for idx in range(len(self)):
            try:
                self[idx]
            except:
                allgood=False
                print(f"couldn't load {self.index.iloc[idx].filename}")
        if allgood:
            print(f"All Good! Loaded {len(self)} images.")

class Deframe(object):
    """check for uniform color boundaries on edges of input and crop them away"""
    
    def __init__(self,aggressive=False,maxPixelFrame=20):
        self.alpha = 0.1 if aggressive else 0.01
        self.maxPixelFrame = maxPixelFrame
    
    def _map2idx(self,frameMap):
        try:
            return frameMap.tolist().index(False)
        except ValueError:
            return self.maxPixelFrame
    
    def _Border(self,img: Tensor):
        """ take greyscale Tensor
            return left,right,top,bottom border size identified """
        top = left = right = bottom = 0
        
        # expected image variance
        hvar,wvar = torch.mean(torch.var(img,dim=0)), torch.mean(torch.var(img,dim=1))
        
        # use image variance and alpha to identify too-uniform frame borders
        top = torch.var(img[:self.maxPixelFrame,:],dim=1) < wvar*(1+self.alpha)
        top = self._map2idx(top)
        
        bottom = torch.var(img[-self.maxPixelFrame:,:],dim=1) < wvar*(1+self.alpha)
        bottom = self._map2idx(bottom)
        
        left = torch.var(img[:,:self.maxPixelFrame],dim=0) < hvar*(1+self.alpha)
        left = self._map2idx(left)
        
        right = torch.var(img[:,-self.maxPixelFrame:],dim=0) < hvar*(1+self.alpha)
        right = self._map2idx(right)
        
        return (top,bottom,right,left)
    
    def __call__(self,img: Tensor):
        top,bottom,right,left = self._Border(torchvision.transforms.Grayscale()(img)[0])
        
        height = img.shape[1]-(top+bottom)
        width  = img.shape[2]-(left+right)
        
        print(f"t{top} b{bottom} l{left} r{right}")
        
        return torchvision.transforms.functional.crop(img,top,left,height,width)