MiniImagenet_task.py

import os
import torch
from torch.utils.data import Dataset
from torchvision.transforms import transforms
import numpy as np
import collections
from PIL import Image
import csv
import random
import json

import pdb

class MiniImagenet(Dataset):
    """
    put mini-imagenet files as :
    root :
        |- images/*.jpg includes all imgeas
        |- train.csv
        |- test.csv
        |- val.csv
    NOTICE: meta-learning is different from general supervised learning, especially the concept of batch and set.
    batch: contains several sets
    sets: conains n_way * k_shot for meta-train set, n_way * n_query for meta-test set.
    """

    def __init__(self, root, mode, batch_size, n_way, k_shot, k_query, resize, task_id, split=1, startidx=0):
        """

        :param root: root path of mini-imagenet
        :param mode: train, val or test
        :param batch_size: batch size of sets, not batch of imgs
        :param n_way:
        :param k_shot:
        :param k_query: num of qeruy imgs per class
        :param resize: resize to
        :param startidx: start to index label from startidx
        """
        self.mode = mode
        self.batch_size = batch_size  # batch of set, not batch of imgs
        self.n_way = n_way  # n-way
        self.k_shot = k_shot  # k-shot
        self.k_query = k_query  # for evaluation
        self.set_size = self.n_way * self.k_shot  # num of samples per set
        self.query_size = self.n_way * self.k_query  # number of samples per set for evaluation
        self.resize = resize  # resize to
        self.split = split
        self.startidx = startidx  # index label not from 0, but from startidx
        print('shuffle DB :%s, b:%d, %d-way, %d-shot, %d-query, resize:%d' % (
        mode, batch_size, n_way, k_shot, k_query, resize))

        if mode == 'train':
            self.transform = transforms.Compose([lambda x: Image.open(x).convert('RGB'),
                                                 transforms.Resize((self.resize, self.resize)),
                                                 #transforms.RandomHorizontalFlip(),
                                                 #transforms.RandomRotation(5),
                                                 transforms.ToTensor(),
                                                 transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
                                                 ])
        else:
            self.transform = transforms.Compose([lambda x: Image.open(x).convert('RGB'),
                                                 transforms.Resize((self.resize, self.resize)),
                                                 transforms.ToTensor(),
                                                 transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
                                                 ])

        self.path = os.path.join(root, 'images')  # image path

        # csvdata is a dict, len(csvdata.keys())=64  64*600=38400
        csvdata = self.loadCSV(os.path.join(root, mode + '.csv'))  # csv path

        self.data = []  # list of list
        self.img2label = {}  # img2label dict
        for i, (k, v) in enumerate(csvdata.items()):
            self.data.append(v)  # [[img1, img2, ...], [img601, ...]]
            self.img2label[k] = i + self.startidx  # {"img_name[:9]":label}
        self.cls_num = len(self.data)  # 64

        ''' Auto maml split train data '''
        if self.split != 1:
            data = list(zip(*self.data))
            data_list = [x for data_tuple in data for x in data_tuple]
            label_list = list(range(self.cls_num)) * len(data)
            self.data_target = list(zip(data_list, label_list))
            data_split = self.data_target[int(self.split[0] * len(data_list)):int(self.split[1] * len(data_list))]
            data_dict = {}
            for v in data_split:
                d = v[0]
                lbl = v[1]
                if lbl in data_dict.keys():
                    data_dict[lbl].append(d)
                else:
                    data_dict[lbl] = [d]

            self.data = []
            for i in range(self.cls_num):
                self.data.append(data_dict[i]) # list of list

        if mode == 'train':
            self.create_batch(self.batch_size)
        else:
            if not os.path.isfile(os.path.join(root, 'test_support_' + str(n_way) + 'way_' + str(k_shot) + 'shot' + '.json')):
                self.create_batch(self.batch_size)
                with open(os.path.join(root, 'test_support_' + str(n_way) + 'way_' + str(k_shot) + 'shot' + '.json'), 'w+') as f_support:
                    json.dump(self.support_x_batch, f_support)
                with open(os.path.join(root, 'test_query_' + str(n_way) + 'way_' + str(k_shot) + 'shot' + '.json'), 'w+') as f_query:
                    json.dump(self.query_x_batch, f_query)

            else:
                self.support_x_batch = json.load(open(os.path.join(root, 'test_support_' + str(n_way) + 'way_' + str(k_shot) + 'shot' + '.json')))
                self.query_x_batch = json.load(open(os.path.join(root, 'test_query_' + str(n_way) + 'way_' + str(k_shot) + 'shot' + '.json')))


            self.support_x_batch = [self.support_x_batch[task_id]]
            self.query_x_batch = [self.query_x_batch[task_id]]



    def loadCSV(self, csvf):
        """
        return a dict saving the information of csv
        :param splitFile: csv file name
        :return: {label:[file1, file2 ...]}
        """
        dictLabels = {}
        with open(csvf) as csvfile:
            csvreader = csv.reader(csvfile, delimiter=',')
            next(csvreader, None)  # skip (filename, label)
            for i, row in enumerate(csvreader):
                filename = row[0]
                label = row[1]
                # append filename to current label
                if label in dictLabels.keys():
                    dictLabels[label].append(filename)
                else:
                    dictLabels[label] = [filename]

        return dictLabels

    def create_batch(self, batch_size):
        """
        create batch for meta-learning.
        ×episode× here means batch, and it means how many sets we want to retain.
        :param episodes: batch size
        :return:
        """

        self.support_x_batch = []  # support set batch
        self.query_x_batch = []  # query set batch
        for b in range(batch_size):  # for each batch
            # 1.select n_way classes randomly, choose 5-way from 64 classes, no duplicate
            selected_cls = np.random.choice(self.cls_num, self.n_way, replace=False)

            np.random.shuffle(selected_cls)
            support_x = []  # list of list
            query_x = []       # list of list
            for cls in selected_cls:
                # 2. select k_shot + k_query for each class

                selected_imgs_idx = np.random.choice(len(self.data[cls]), self.k_shot + self.k_query, replace=False)
                np.random.shuffle(selected_imgs_idx)
                indexDtrain = np.array(selected_imgs_idx[:self.k_shot])  # idx for Dtrain
                indexDtest = np.array(selected_imgs_idx[self.k_shot:])  # idx for Dtest
                support_x.append(
                    np.array(self.data[cls])[indexDtrain].tolist())  # get all images filename for current Dtrain
                query_x.append(np.array(self.data[cls])[indexDtest].tolist())

            # shuffle the correponding relation between support set and query set
            random.shuffle(support_x)  # 5*1
            random.shuffle(query_x)    # 5*15

            self.support_x_batch.append(support_x)  # append set to current sets, 10000*5*1
            self.query_x_batch.append(query_x)  # append sets to current sets, 10000*5*15



    def __getitem__(self, index):
        """
        index means index of sets, 0<= index <= batch_size-1
        :param index:
        :return:
        """
        # [set_size, 3, resize, resize]
        support_x = torch.FloatTensor(self.set_size, 3, self.resize, self.resize)
        # [set_size]
        #support_y = np.zeros((self.set_size), dtype=np.int)
        # [query_size, 3, resize, resize]
        query_x = torch.FloatTensor(self.query_size, 3, self.resize, self.resize)
        # [query_size]
        #query_y = np.zeros((self.query_size), dtype=np.int)

        flatten_support_x = []  # list of path
        #print(index)
        for sublist in self.support_x_batch[index]:
            for item in sublist:
                flatten_support_x.append(os.path.join(self.path, item))

        # flatten_support_x = [os.path.join(self.path, item)
        #                      for sublist in self.support_x_batch[index] for item in sublist]
        support_y = []
        for sublist in self.support_x_batch[index]:
            for item in sublist:
                support_y.append(self.img2label[item[:9]])
        support_y = np.array(support_y)
        # support_y = np.array(
        #     [self.img2label[item[:9]]  # filename:n0153282900000005.jpg, the first 9 characters treated as label
        #      for sublist in self.support_x_batch[index] for item in sublist]).astype(np.int32)

        flatten_query_x = [os.path.join(self.path, item)
                           for sublist in self.query_x_batch[index] for item in sublist]
        query_y = np.array([self.img2label[item[:9]]
                            for sublist in self.query_x_batch[index] for item in sublist]).astype(np.int32)

        # print('global:', support_y, query_y)
        # support_y: [set_size]
        # query_y: [query_size]
        # unique: [n-way], sorted
        unique = np.unique(support_y)
        random.shuffle(unique)
        # relative means the label ranges from 0 to n-way

        support_y_relative = np.zeros(self.set_size)  # 5*1
        query_y_relative = np.zeros(self.query_size)  # 5*15
        for idx, label in enumerate(unique):
            support_y_relative[support_y == label] = idx
            query_y_relative[query_y == label] = idx

        # print('relative:', support_y_relative, query_y_relative)

        for i, path in enumerate(flatten_support_x):
            support_x[i] = self.transform(path)

        for i, path in enumerate(flatten_query_x):
            query_x[i] = self.transform(path)
        # print(support_set_y)
        # return support_x, torch.LongTensor(support_y), query_x, torch.LongTensor(query_y)
        return support_x, torch.LongTensor(support_y_relative), query_x, torch.LongTensor(query_y_relative)

    def __len__(self):
        # as we have built up to batch_size of sets, you can sample some small batch size of sets.
        return len(self.support_x_batch)


if __name__ == '__main__':
    # the following episode is to view one set of images via tensorboard.
    from torchvision.utils import make_grid
    from matplotlib import pyplot as plt
    from tensorboardX import SummaryWriter
    import time

    plt.ion()

    tb = SummaryWriter('runs', 'mini-imagenet')
    mini = MiniImagenet('../mini-imagenet/', mode='train', n_way=5, k_shot=1, k_query=1, batch_size=1000, resize=168)

    for i, set_ in enumerate(mini):
        # support_x: [k_shot*n_way, 3, 84, 84]
        support_x, support_y, query_x, query_y = set_

        support_x = make_grid(support_x, nrow=2)
        query_x = make_grid(query_x, nrow=2)

        plt.figure(1)
        plt.imshow(support_x.transpose(2, 0).numpy())
        plt.pause(0.5)
        plt.figure(2)
        plt.imshow(query_x.transpose(2, 0).numpy())
        plt.pause(0.5)

        tb.add_image('support_x', support_x)
        tb.add_image('query_x', query_x)

        time.sleep(5)

    tb.close()