File size: 9,954 Bytes
2daf4c3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
import os
from PIL import Image
import torch
import torch.nn as nn
from ctc_decoder import best_path, beam_search
import pytorch_lightning as pl
from torchmetrics import CharErrorRate
from torchvision.transforms import Compose, Resize, Grayscale, ToTensor
from torchvision.utils import make_grid

from modelling import HandwritingRecognition


class HandwritingRecogTrainModule(pl.LightningModule):
    def __init__(self, hparams, index_to_labels, label_to_index):
        super().__init__()
        # save_hyperparameters saves the parameters in the signature
        self.save_hyperparameters()
        self.chars = ' -ABCDEFGHIJKLMNOPQRSTUVWXYZ'
        self.model = HandwritingRecognition(self.hparams['hparams']['gru_input_size'], self.hparams['hparams']['gru_hidden_size'],
                                            self.hparams['hparams']['gru_num_layers'], self.hparams['hparams']['num_classes'])
        self.criterion = nn.CTCLoss(blank=28, zero_infinity=True, reduction='mean')
        self.transforms = Compose([Resize((self.hparams['hparams']['input_height'], self.hparams['hparams']['input_width'])), Grayscale(),
                                                              ToTensor()])
        self.char_metric = CharErrorRate()

    def forward(self, the_image):
        out = self.model(the_image)
        out = out.permute(1, 0, 2)
        out = torch.exp(out)
        return out

    def intermediate_operation(self, batch):
        transformed_images = batch['transformed_images']
        labels = batch['labels']
        target_lens = batch['target_lens']

        output = self.model(transformed_images)

        N = output.size(1)
        input_length = output.size(0)
        input_lengths = torch.full(size=(N,), fill_value=input_length, dtype=torch.int32)

        loss = self.criterion(output, labels, input_lengths, target_lens)
        return loss, output

    def training_step(self, batch, batch_idx):
        loss, preds = self.intermediate_operation(batch)
        with torch.inference_mode():
            preds = preds.permute(1, 0, 2)
            preds = torch.exp(preds)
            ground_truth = batch['labels']
            target_lens = batch['target_lens']
            ground_truth = ground_truth.cpu().detach().numpy()
            target_lens = target_lens.cpu().detach().numpy()
            preds = preds.cpu().detach().numpy()
            actual_predictions = []
            for pred in preds:
                actual_predictions.append(best_path(pred, self.chars))
            exact_matches = 0
            actual_ground_truths = []
            for i, predicted_string in enumerate(actual_predictions):
                ground_truth_sample = ground_truth[i][0:target_lens[i]]
                ground_truth_string = [self.hparams.index_to_labels[index] for index in ground_truth_sample]
                ground_truth_string = ''.join(ground_truth_string)
                actual_ground_truths.append(ground_truth_string)
                if predicted_string == ground_truth_string:
                    exact_matches += 1
            exact_match_percentage = (exact_matches / len(preds)) * 100
            char_error_rate = self.char_metric(actual_predictions, actual_ground_truths)
            self.log_dict({'train-loss': loss, 'train-exact-match': exact_match_percentage,
                           'train-char_error_rate': char_error_rate}, prog_bar=True, on_epoch=True, on_step=False)
        return loss

    def validation_step(self, batch, batch_idx):
        loss, preds = self.intermediate_operation(batch)
        preds = preds.permute(1, 0, 2)
        preds = torch.exp(preds)
        ground_truth = batch['labels']
        target_lens = batch['target_lens']
        ground_truth = ground_truth.cpu().detach().numpy()
        target_lens = target_lens.cpu().detach().numpy()
        preds = preds.cpu().detach().numpy()
        actual_predictions = []
        for pred in preds:
            actual_predictions.append(best_path(pred, self.chars))
        exact_matches = 0
        actual_ground_truths = []
        for i, predicted_string in enumerate(actual_predictions):
            ground_truth_sample = ground_truth[i][0:target_lens[i]]
            ground_truth_string = [self.hparams.index_to_labels[index] for index in ground_truth_sample]
            ground_truth_string = ''.join(ground_truth_string)
            actual_ground_truths.append(ground_truth_string)
            if predicted_string == ground_truth_string:
                exact_matches += 1
        char_error_rate = self.char_metric(actual_predictions, actual_ground_truths)
        exact_match_percentage = (exact_matches / len(preds)) * 100
        if batch_idx % self.trainer.num_val_batches[0] == 0:
            small_batch = batch['transformed_images'][0:16]
            small_batch_predictions = actual_predictions[0:16]
            captions = small_batch_predictions
            sampled_img_grid = make_grid(small_batch)
            self.logger.log_image('Sample_Images', [sampled_img_grid], caption=[str(captions)])

        self.log_dict({'val-loss': loss, 'val-exact-match': exact_match_percentage,
                       'val-char-error-rate': char_error_rate}, prog_bar=False, on_epoch=True, on_step=False)

    def configure_optimizers(self):
        return torch.optim.Adam(self.parameters(), lr=self.hparams['hparams']['lr'])


def convert_to_torchscript(trained_path):
    hparams = {
        'train_img_path': './data/kaggle-handwriting-recognition/train_v2/train/',
        'lr': 1e-3, 'val_img_path': './data/kaggle-handwriting-recognition/validation_v2/validation/',
        'test_img_path': './data/kaggle-handwriting-recognition/test_v2/test/',
        'data_path': './data/kaggle-handwriting-recognition', 'gru_input_size': 256,
        'train_batch_size': 64, 'val_batch_size': 256, 'input_height': 36, 'input_width': 324, 'gru_hidden_size': 128,
        'gru_num_layers': 1, 'num_classes': 28
    }
    index_to_labels = {0: ' ', 1: '-', 2: 'A', 3: 'B', 4: 'C', 5: 'D', 6: 'E', 7: 'F', 8: 'G', 9: 'H', 10: 'I',
                       11: 'J', 12: 'K', 13: 'L', 14: 'M', 15: 'N', 16: 'O', 17: 'P', 18: 'Q', 19: 'R', 20: 'S',
                       21: 'T', 22: 'U', 23: 'V', 24: 'W', 25: 'X', 26: 'Y', 27: 'Z'}
    label_to_index = {' ': 0, '-': 1, 'A': 2, 'B': 3, 'C': 4, 'D': 5, 'E': 6, 'F': 7, 'G': 8, 'H': 9, 'I': 10, 'J': 11,
                      'K': 12, 'L': 13, 'M': 14, 'N': 15, 'O': 16, 'P': 17, 'Q': 18, 'R': 19, 'S': 20, 'T': 21, 'U': 22,
                      'V': 23, 'W': 24, 'X': 25, 'Y': 26, 'Z': 27}
    model = HandwritingRecogTrainModule(hparams, index_to_labels, label_to_index)
    script = model.to_torchscript()
    print("The script:", script)
    torch.jit.save(script, './final-models/torchscript-model/handwritten-name_new.pt')

def test_model():
    pl.seed_everything(2564)
    hparams = {
        'train_img_path': './data/kaggle-handwriting-recognition/train_v2/train/',
        'lr': 1e-3, 'val_img_path': './data/kaggle-handwriting-recognition/validation_v2/validation/',
        'test_img_path': './data/kaggle-handwriting-recognition/test_v2/test/',
        'data_path': './data/kaggle-handwriting-recognition', 'gru_input_size': 256,
        'train_batch_size': 64, 'val_batch_size': 256, 'input_height': 36, 'input_width': 324, 'gru_hidden_size': 128,
        'gru_num_layers': 1, 'num_classes': 28
    }
    index_to_labels = {0: ' ', 1: '-', 2: 'A', 3: 'B', 4: 'C', 5: 'D', 6: 'E', 7: 'F', 8: 'G', 9: 'H', 10: 'I',
                       11: 'J', 12: 'K', 13: 'L', 14: 'M', 15: 'N', 16: 'O', 17: 'P', 18: 'Q', 19: 'R', 20: 'S',
                       21: 'T', 22: 'U', 23: 'V', 24: 'W', 25: 'X', 26: 'Y', 27: 'Z'}
    label_to_index = {' ': 0, '-': 1, 'A': 2, 'B': 3, 'C': 4, 'D': 5, 'E': 6, 'F': 7, 'G': 8, 'H': 9, 'I': 10, 'J': 11,
                      'K': 12, 'L': 13, 'M': 14, 'N': 15, 'O': 16, 'P': 17, 'Q': 18, 'R': 19, 'S': 20, 'T': 21, 'U': 22,
                      'V': 23, 'W': 24, 'X': 25, 'Y': 26, 'Z': 27}

    model = HandwritingRecogTrainModule.load_from_checkpoint(
        './lightning_logs/CNNR_run_new_version/108xqa9y/checkpoints/'
        'epoch=21-val-loss=0.206-val-exact-match=81.46109771728516-val-char-error-rate=0.04727236181497574.ckpt')
    input_image = Image.open(os.path.join(hparams['train_img_path'], 'TRAIN_96628.jpg'))
    output = model(input_image)
    print(output)

def test_inference():
    transforms = Compose([Resize((36, 324)), Grayscale(), ToTensor()])
    input_image = Image.open(os.path.join('./data/kaggle-handwriting-recognition/train_v2/train/', 'TRAIN_96628.jpg'))
    transformed_image = transforms(input_image)
    # path = './lightning_logs/CNNR_run_64_2grulayers_0.3dropout/3182ng3f/checkpoints'
    # model_weights = 'epoch=47-val-loss=0.190-val-exact-match=83.1511001586914-val-char-error-rate=0.042957037687301636.ckpt'
    # trained_path = os.path.join(path, model_weights)
    # model = HandwritingRecogTrainModule.load_from_checkpoint(trained_path)
    # transformed_image = torch.unsqueeze(transformed_image, 0)
    # model.eval()
    # out = model(transformed_image)
    script_path = './final-models/torchscript-model/handwritten-name_new.pt'
    scripted_module = torch.jit.load(script_path)
    out = scripted_module(transformed_image)
    print("The final out shape:", out.shape)
    print("The final out is :", out)
    out = out.cpu().detach().numpy()
    chars = ' -ABCDEFGHIJKLMNOPQRSTUVWXYZ'
    for sample in out:
        predicted_string = beam_search(sample, chars, beam_width=2)
    print(predicted_string)

def test_convert_to_torchscript():
    path = './lightning_logs/CNNR_run_new_version/108xqa9y/checkpoints/'
    model_weights = 'epoch=21-val-loss=0.206-val-exact-match=81.46109771728516-val-char-error-rate=0.04727236181497574.ckpt'
    trained_path = os.path.join(path, model_weights)
    convert_to_torchscript(trained_path)


if __name__ == '__main__':
    test_inference()