Source code for domid.tasks.task_wsi

import numpy as np
import pandas as pd
from domainlab.tasks.utils_task import ImSize
from domainlab.utils.utils_classif import mk_dummy_label_list_str
from torch.utils.data import random_split
from torchvision import transforms

from domid.dsets.dset_wsi import DsetWSI
from domid.tasks.b_task_cluster import NodeTaskDictCluster


class NodeTaskWSI(NodeTaskDictCluster):
    @property
    def list_str_y(self):
        """
        WSI task has no labels because it's used for unsupervised learning (the available annotations are considered domains instead)
        """
        return mk_dummy_label_list_str("dummy", 6)

    @property
    def isize(self):
        """
        :return: image size object storing image channels, height, width.
        """
        return ImSize(3, 64, 64)

    def get_list_domains(self):
        """
        Get list of domain names

        :return: list of domain names
        """
        return mk_dummy_label_list_str("digit", 313)  # FIXME: "digit" is wrong for WSI

    def get_dset_by_domain(self, args, na_domain, split=True):
        """Get a dataset by 'domain'

        :param args: command line arguments
        :param na_domain: domain name
        :param split: whether a training/validation split is performed (the
            training split portion will be determined by args.split); for test
            set, no need to split; args.split: by default, split is set to be
            zero which in python can be evaluated in if statement, in which case,
            no separate validation set will be created. Otherwise, this argument
            is the percentage of the data to be used as training set, while the
            rest will be used as validation set.
        :return: training dataset, validation dataset
        """
        ratio_split = float(args.split) if split else False
        # by default, split is set to be zero which in python can
        # be evaluated in if statement, in which case, no validation
        # set will be created. Otherwise, this argument is
        # the split ratio
        ind_global = self.get_list_domains().index(na_domain)
        df = pd.read_csv(args.meta_data_csv)
        mask = df[df["subject"] == ind_global]
        # mask = df['resp'].values == ind_global  # response (0 o 1)
        img_paths = np.array(mask["path"])  # [:400]

        trans = [transforms.Resize((64, 64))]  # [transforms.Resize((self.ImSize.w, self.ImSize.h))]
        dset = DsetWSI(class_num=ind_global, path=img_paths, args=args, transform=trans)

        # split dset into training and validation sets
        if ratio_split > 0:
            train_len = int(len(dset) * ratio_split)
            val_len = len(dset) - train_len
            train_set, val_set = random_split(dset, [train_len, val_len])
        else:
            train_set = dset
            val_set = dset
        return train_set, val_set