-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocessing.py
218 lines (178 loc) · 7.34 KB
/
preprocessing.py
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
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
import torch
from torch.utils.data import Dataset, DataLoader
from torch.nn.utils.rnn import pad_sequence
from torchvision import transforms
from torchvision.models import mobilenet_v2, MobileNet_V2_Weights
from PIL import Image
import numpy as np
import matplotlib.pyplot as plt
import random
import os
import pickle
import argparse
import string
import tiktoken
import logging
# Set up logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
# max xml length = 994
class IAMDataset(Dataset):
def __init__(self, data):
self.data = data
def __len__(self):
return len(self.data)
def __getitem__(self, idx):
strokes, text, image = self.data[idx]
strokes = torch.Tensor(strokes).to(dtype=torch.float32)
text = torch.Tensor(text).to(dtype=torch.int64)
image = torch.Tensor(image).permute(2, 0, 1).to(dtype=torch.uint8) # Change from (H, W, C) to (C, H, W)
return strokes, text, image
def remove_whitespace(img, thresh, remove_middle=False):
row_mins = np.amin(img, axis=1)
col_mins = np.amin(img, axis=0)
rows = np.where(row_mins < thresh)
cols = np.where(col_mins < thresh)
if remove_middle:
return img[rows[0]][:, cols[0]]
else:
rows, cols = rows[0], cols[0]
return img[rows[0]:rows[-1], cols[0]:cols[-1]]
def norms(x):
return np.linalg.norm(x, axis=-1)
def combine_strokes(x, n):
s, s_neighbors = x[::2, :2], x[1::2, :2]
if len(x) % 2 != 0:
s = s[:-1]
values = norms(s) + norms(s_neighbors) - norms(s + s_neighbors)
ind = np.argsort(values)[:n]
x[ind*2] += x[ind*2+1]
x[ind*2, 2] = np.greater(x[ind*2, 2], 0)
x = np.delete(x, ind*2+1, axis=0)
x[:, :2] /= np.std(x[:, :2])
return x
def parse_page_text(dir_path, id):
dict = {}
with open(os.path.join(dir_path, id)) as f:
has_started = False
line_num = -1
for l in f:
if 'CSR' in l:
has_started = True
if has_started:
if line_num > 0:
dict[f"{id[:-4]}-{line_num:02d}"] = l.strip()
line_num += 1
logger.info(f"Parsed page text for {id}, found {len(dict)} lines")
return dict
def create_dict(path):
dict = {}
for dir in os.listdir(path):
if dir != '.DS_Store':
dirpath = os.path.join(path, dir)
for subdir in os.listdir(dirpath):
if subdir != '.DS_Store':
subdirpath = os.path.join(dirpath, subdir)
forms = os.listdir(subdirpath)
for f in forms:
dict.update(parse_page_text(subdirpath, f))
logger.info(f"Created dictionary with {len(dict)} entries")
return dict
def parse_stroke_xml(path): # pad the xmls to 1000 strokes so everything is consistent
with open(path) as xml:
xml = xml.readlines()
strokes = []
previous = None
for i, l in enumerate(xml):
if 'Point' in l:
x_ind, y_ind, y_end = l.index('x='), l.index('y='), l.index('time=')
x = int(l[x_ind+3:y_ind-2])
y = int(l[y_ind+3:y_end-2])
is_end = 1.0 if '/Stroke' in xml[i+1] else 0.0
if previous is None:
previous = [x, -y]
else:
strokes.append([x - previous[0], -y - previous[1], is_end])
previous = [x, -y]
strokes = np.array(strokes)
strokes[:, 2] = np.roll(strokes[:, 2], 1)
strokes[:, :2] /= np.std(strokes[:, :2])
for i in range(3):
strokes = combine_strokes(strokes, int(len(strokes)*0.2))
logger.info(f"Parsed stroke XML for {path}, found {len(strokes)} strokes")
return strokes
def read_img(path, height):
img = Image.open(path)
img_arr = np.array(img)
img_arr = remove_whitespace(img_arr, thresh=127)
h, w = img_arr.shape
new_w = height * w // h # Use integer division
img_resized = Image.fromarray(img_arr).resize((new_w, height), Image.BILINEAR)
return torch.Tensor(np.array(img_resized).astype('uint8'))
def create_dataset(formlist, strokes_path, images_path, tokenizer, text_dict, height): # max sentence length is 24
dataset = []
with open(formlist) as f:
forms = f.readlines()
for f in forms:
path = os.path.join(strokes_path, f[1:4], f[1:8])
offline_path = os.path.join(images_path, f[1:4], f[1:8])
samples = [s for s in os.listdir(path) if f[1:-1] in s]
offline_samples = [s for s in os.listdir(offline_path) if f[1:-1] in s]
shuffled_offline_samples = offline_samples.copy()
random.shuffle(shuffled_offline_samples)
for i in range(len(samples)):
sample_id = samples[i][:-4]
if sample_id not in text_dict:
logger.warning(f"Sample {sample_id} not found in text dictionary")
continue
"""data to be added (this is here for debugging)"""
stroke_vec = torch.Tensor(parse_stroke_xml(os.path.join(path, samples[i])))
tokenized_string = torch.Tensor(tokenizer.encode(text_dict[sample_id]))
img_vec = read_img(os.path.join(offline_path, shuffled_offline_samples[i]), height)
dataset.append((
stroke_vec,
tokenized_string,
img_vec
))
logger.info(f"Created dataset with {len(dataset)} samples")
return dataset
def main():
parser = argparse.ArgumentParser()
parser.add_argument('-t', '--text_path', default='./data/ascii', help='path to text labels, default ./data/ascii')
parser.add_argument('-s', '--strokes_path', default='./data/lineStrokes',
help='path to stroke xml, default ./data/lineStrokes')
parser.add_argument('-i', '--images_path', default='./data/lineImages',
help='path to line images, default ./data/lineImages')
parser.add_argument('-H', '--height', type=int, default=96,
help='the height of offline images, default 96')
args = parser.parse_args()
t_path = args.text_path
s_path = args.strokes_path
i_path = args.images_path
H = args.height
train_info = './data/trainset.txt'
val1_info = './data/testset_f.txt'
val2_info = './data/testset_t.txt'
test_info = './data/testset_v.txt'
# Initialize tiktoken tokenizer
tokenizer = tiktoken.get_encoding("p50k_base")
labels = create_dict(t_path)
train_strokes = create_dataset(train_info, s_path, i_path, tokenizer, labels, H)
val1_strokes = create_dataset(val1_info, s_path, i_path, tokenizer, labels, H)
val2_strokes = create_dataset(val2_info, s_path, i_path, tokenizer, labels, H)
test_strokes = create_dataset(test_info, s_path, i_path, tokenizer, labels, H)
train_strokes += val1_strokes
train_strokes += val2_strokes
random.shuffle(train_strokes)
random.shuffle(test_strokes)
# Create datasets
train_dataset = IAMDataset(train_strokes)
test_dataset = IAMDataset(test_strokes)
# Save the datasets if needed
with open('./data/train_strokes.p', 'wb') as f:
pickle.dump(train_strokes, f)
with open('./data/test_strokes.p', 'wb') as f:
pickle.dump(test_strokes, f)
logger.info("Datasets created and saved successfully")
if __name__ == '__main__':
main()