image_captioning/utils_torch.py at main · elyaishere/image_captioning · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
import itertools
import os

import numpy as np
import torch
from PIL import Image
from nltk.translate.bleu_score import sentence_bleu
from tqdm.auto import tqdm

def check_create_dir(full_path):
    if not os.path.isdir(full_path):
        os.makedirs(full_path)

def get_picture_caption(idx, dataset, model, idx2word, beam_width=-1):
    """
    GREEDY! beam_width is not used!!!
    """
    im, cp, _ = dataset[idx]
    capidx = model.sample(im.unsqueeze(0))[0].detach().cpu().numpy()
    caption_pred = ''.join(list(itertools.takewhile(lambda word: word.strip() != '<end>',
                                                         map(lambda idx: idx2word[idx]+' ', iter(capidx))))[1:])
    return caption_pred


def greedy_predictions_gen(encoding_dict, model, word2idx, idx2word, images, max_len,
                           startseq="<start>", endseq="<end>", device=torch.device('cpu')):
    def greedy_search_predictions_util(image):
        start_word = [startseq]
        with torch.no_grad():
            while True:
                par_caps = torch.LongTensor([word2idx[i] for i in start_word])
                par_caps = padding_tensor([par_caps], maxlen=max_len).to(device=device)
                e = encoding_dict[image[len(images):]].unsqueeze(0)
                preds = model(e, par_caps).cpu().numpy()
                word_pred = idx2word[np.argmax(preds[0])]  # [0] is for first elm of batch
                start_word.append(word_pred)

                if word_pred == endseq or len(start_word) > max_len:
                    break
        return ' '.join(start_word[1:-1])

    return greedy_search_predictions_util


def beam_search_predictions_gen(beam_index, encoding_dict, model, word2idx, idx2word, images, max_len,
                                startseq="<start>", endseq="<end>", device=torch.device('cpu')):
    def beam_search_predictions_util(image):
        start = [word2idx[startseq]]

        start_word = [[start, 0.0]]

        while len(start_word[0][0]) < max_len:
            temp = []
            for s in start_word:
                with torch.no_grad():
                    par_caps = torch.LongTensor(s[0])
                    par_caps = padding_tensor([par_caps], maxlen=max_len).to(device=device)
                    e = encoding_dict[image[len(images):]].unsqueeze(0)
                    preds = model(e, par_caps).cpu().numpy()

                    word_preds = np.argsort(preds[0])[-beam_index:]

                    # Getting the top <beam_index>(n) predictions and creating a
                    # new list so as to put them via the model again
                    for w in word_preds:
                        next_cap, prob = s[0][:], s[1]
                        next_cap.append(w)
                        prob += preds[0][w]
                        temp.append([next_cap, prob])

            start_word = temp
            # Sorting according to the probabilities
            start_word = sorted(start_word, reverse=False, key=lambda l: l[1])
            # Getting the top words
            start_word = start_word[-beam_index:]

        start_word = start_word[-1][0]
        intermediate_caption = [idx2word[i] for i in start_word]

        final_caption = []

        for i in intermediate_caption:
            if i != endseq:
                final_caption.append(i)
            else:
                break

        final_caption = ' '.join(final_caption[1:])
        return final_caption

    return beam_search_predictions_util


def split_data(l, img, images):
    temp = []
    for i in img:
        if i[len(images):] in l:
            temp.append(i)
    return temp


def get_bleu_score(img_to_caplist_dict, caption_gen_func, device=torch.device('cpu')):
    bleu_score = 0.0
    for k, v in tqdm(img_to_caplist_dict.items()):
        candidate = caption_gen_func(k).split()
        references = [s.split() for s in v]
        bleu_score += sentence_bleu(references, candidate)
    return bleu_score / len(img_to_caplist_dict)


def print_eval_metrics(img_cap_dict, encoding_dict, model, word2idx, idx2word, images, max_len,
                       device=torch.device('cpu')):
    print('\t\tGreedy:            ',
          get_bleu_score(img_cap_dict, greedy_predictions_gen(encoding_dict=encoding_dict, model=model,
                                                              word2idx=word2idx, idx2word=idx2word,
                                                              images=images, max_len=max_len)))
    for k in [3, 5, 7]:
        print(f'\t\tBeam Search k={k}:', get_bleu_score(img_cap_dict,
                                                        beam_search_predictions_gen(beam_index=k,
                                                                                    encoding_dict=encoding_dict,
                                                                                    model=model,
                                                                                    word2idx=word2idx,
                                                                                    idx2word=idx2word,
                                                                                    images=images, max_len=max_len)))


def padding_tensor(sequences, maxlen):
    """
    :param sequences: list of tensors
    :param maxlen: fixed length of output tensors
    :return:
    """
    num = len(sequences)
    # max_len = max([s.size(0) for s in sequences])
    out_dims = (num, maxlen)
    out_tensor = sequences[0].data.new(*out_dims).fill_(0)
    for i, tensor in enumerate(sequences):
        length = tensor.size(0)
        out_tensor[i, :length] = tensor
    return out_tensor


def words_from_tensors_fn(idx2word, max_len=40, startseq='<start>', endseq='<end>'):
    def words_from_tensors(captions: np.array) -> list:
        """
        :param captions: [b, max_len]
        :return:
        """
        captoks = []
        for capidx in captions:
            # capidx = [1, max_len]
            captoks.append(list(itertools.takewhile(lambda word: word != endseq,
                                                    map(lambda idx: idx2word[idx], iter(capidx))))[1:])
        return captoks

    return words_from_tensors

def get_best_and_worst_quality_captions(dset, model, idx2word, metric):
    maxs = 0
    mins = 100
    idmint = None
    idmaxt = None
    for idx in range(len(dset)):
        generated_caption = get_picture_caption(idx, dset, model, idx2word)
        score = metric([[i.split() for i in dset.get_image_captions(idx)[1]]], [generated_caption.split()], n=4)
        if maxs < score:
            maxs = score
            idmaxt = idx
        if mins > score:
            mins = score
            idmint = idx
    return [idmaxt, idmint]