Update dataset.py
Browse filesupdates for split folder and removed local dir
- dataset.py +37 -64
dataset.py
CHANGED
@@ -1,86 +1,88 @@
|
|
1 |
-
|
2 |
import os
|
3 |
import json
|
4 |
-
import time
|
5 |
from PIL import Image
|
6 |
from torch.utils.data import Dataset, DataLoader
|
7 |
from torchvision import transforms
|
8 |
|
9 |
class RQADataset(Dataset):
|
10 |
-
def __init__(self,
|
11 |
"""
|
12 |
Initializes the dataset.
|
13 |
|
14 |
Args:
|
15 |
-
|
|
|
16 |
transform: Optional transform to be applied on a sample.
|
17 |
"""
|
18 |
-
self.
|
19 |
-
self.
|
20 |
-
self.filter_list_file = data_config.filter_list
|
21 |
-
self.train = data_config.train
|
22 |
self.transform = transform or transforms.Compose([
|
23 |
-
transforms.Resize((512, 512))
|
|
|
24 |
])
|
25 |
|
|
|
26 |
self.questions = []
|
27 |
-
|
28 |
-
# Load file names for testing or use all files for training
|
29 |
self.file_names = self._load_file_names()
|
30 |
self._create_questions()
|
31 |
print(f"Total Questions Loaded: {len(self.questions)}")
|
32 |
|
33 |
def _load_file_names(self):
|
34 |
"""
|
35 |
-
Loads the list of file names to be processed.
|
36 |
|
37 |
Returns:
|
38 |
A list of file names without extensions.
|
39 |
"""
|
40 |
-
if
|
41 |
-
|
|
|
|
|
42 |
file_names = [line.strip() for line in f]
|
43 |
-
print(f"Loaded {len(file_names)} test files from {
|
44 |
-
return file_names
|
45 |
else:
|
46 |
-
#
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
def _create_questions(self):
|
50 |
"""
|
51 |
Creates the list of questions from JSON files.
|
52 |
"""
|
53 |
-
start_time = time.time()
|
54 |
unused_count = 0
|
55 |
-
|
56 |
for file_name in self.file_names:
|
57 |
-
|
58 |
-
if
|
59 |
-
|
60 |
-
|
61 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
62 |
with open(json_path, 'r') as f:
|
63 |
json_data = json.load(f)
|
64 |
for item in json_data:
|
65 |
if 'PMC_ID' not in item or 'qa_id' not in item:
|
66 |
continue # Ensure all necessary fields are present
|
67 |
-
item['image_path'] = os.path.join(self.img_dir, item['PMC_ID']
|
68 |
if os.path.exists(item['image_path']):
|
69 |
self.questions.append(item)
|
70 |
else:
|
71 |
unused_count += 1
|
72 |
-
|
73 |
-
elapsed_time = time.time() - start_time
|
74 |
-
print(f"Elapsed time to create questions: {elapsed_time:.2f} seconds = {elapsed_time/60:.2f} minutes")
|
75 |
-
print(f'Total unused/used images: {unused_count} / {len(self.file_names) - unused_count}')
|
76 |
|
77 |
def __len__(self):
|
78 |
return len(self.questions)
|
79 |
|
80 |
def __getitem__(self, idx):
|
81 |
-
return self._load_data(idx)
|
82 |
-
|
83 |
-
def _load_data(self, idx):
|
84 |
"""
|
85 |
Loads a single data point.
|
86 |
|
@@ -132,37 +134,8 @@ class RQADataset(Dataset):
|
|
132 |
}
|
133 |
|
134 |
if __name__ == "__main__":
|
135 |
-
#
|
136 |
-
|
137 |
-
img_dir = '/home/jupyter/RealCQA/code/data/RQA_V0/images'
|
138 |
-
json_dir = '/home/jupyter/RealCQA/code/data/RQA_V0/qa'
|
139 |
-
filter_list = '/home/jupyter/RealCQA/code/data/RQA_V0/test_filenames.txt'
|
140 |
-
train = False # Set to False to prepare the test files
|
141 |
-
|
142 |
-
# Initialize dataset
|
143 |
-
dataset = RQADataset(DataConfig)
|
144 |
-
|
145 |
-
# Test loading a single item
|
146 |
-
print(f"Number of samples in dataset: {len(dataset)}")
|
147 |
-
sample = dataset[0]
|
148 |
-
print("Sample data:", sample)
|
149 |
-
|
150 |
-
# Initialize DataLoader
|
151 |
-
dataloader = DataLoader(dataset, batch_size=4, collate_fn=RQADataset.custom_collate)
|
152 |
-
|
153 |
-
# Test DataLoader
|
154 |
-
for batch in dataloader:
|
155 |
-
print("Batch data:", batch)
|
156 |
-
break # Load only one batch for testing
|
157 |
-
|
158 |
-
class DataConfig:
|
159 |
-
img_dir = '/home/jupyter/RealCQA/code/data/RQA_V0/images'
|
160 |
-
json_dir = '/home/jupyter/RealCQA/code/data/RQA_V0/qa'
|
161 |
-
filter_list = '/home/jupyter/RealCQA/code/data/RQA_V0/test_filenames.txt'
|
162 |
-
train = True # Set to False to prepare the test files
|
163 |
-
|
164 |
-
# Initialize dataset
|
165 |
-
dataset = RQADataset(DataConfig)
|
166 |
|
167 |
# Test loading a single item
|
168 |
print(f"Number of samples in dataset: {len(dataset)}")
|
|
|
|
|
1 |
import os
|
2 |
import json
|
|
|
3 |
from PIL import Image
|
4 |
from torch.utils.data import Dataset, DataLoader
|
5 |
from torchvision import transforms
|
6 |
|
7 |
class RQADataset(Dataset):
|
8 |
+
def __init__(self, data_dir, split='train', transform=None):
|
9 |
"""
|
10 |
Initializes the dataset.
|
11 |
|
12 |
Args:
|
13 |
+
data_dir: Base directory of the dataset on the Hugging Face Hub.
|
14 |
+
split: Split of the dataset ('train' or 'test').
|
15 |
transform: Optional transform to be applied on a sample.
|
16 |
"""
|
17 |
+
self.data_dir = data_dir
|
18 |
+
self.split = split
|
|
|
|
|
19 |
self.transform = transform or transforms.Compose([
|
20 |
+
transforms.Resize((512, 512)),
|
21 |
+
transforms.ToTensor()
|
22 |
])
|
23 |
|
24 |
+
# Initialize lists to hold image and question data
|
25 |
self.questions = []
|
|
|
|
|
26 |
self.file_names = self._load_file_names()
|
27 |
self._create_questions()
|
28 |
print(f"Total Questions Loaded: {len(self.questions)}")
|
29 |
|
30 |
def _load_file_names(self):
|
31 |
"""
|
32 |
+
Loads the list of file names to be processed based on the split.
|
33 |
|
34 |
Returns:
|
35 |
A list of file names without extensions.
|
36 |
"""
|
37 |
+
if self.split == 'test':
|
38 |
+
# Load test file names from the list provided on Hugging Face
|
39 |
+
filter_list_file = os.path.join(self.data_dir, 'test_filenames.txt')
|
40 |
+
with open(filter_list_file, 'r') as f:
|
41 |
file_names = [line.strip() for line in f]
|
42 |
+
print(f"Loaded {len(file_names)} test files from {filter_list_file}")
|
|
|
43 |
else:
|
44 |
+
# For training, use all JSON files from all directories
|
45 |
+
file_names = []
|
46 |
+
for json_dir in ['jsons', 'jsons2', 'jsons3']:
|
47 |
+
json_dir_path = os.path.join(self.data_dir, json_dir)
|
48 |
+
json_files = [os.path.splitext(file)[0] for file in os.listdir(json_dir_path) if file.endswith('.json')]
|
49 |
+
file_names.extend(json_files)
|
50 |
+
return file_names
|
51 |
|
52 |
def _create_questions(self):
|
53 |
"""
|
54 |
Creates the list of questions from JSON files.
|
55 |
"""
|
|
|
56 |
unused_count = 0
|
|
|
57 |
for file_name in self.file_names:
|
58 |
+
# Determine which folder contains the current JSON file
|
59 |
+
if file_name in os.listdir(os.path.join(self.data_dir, 'jsons')):
|
60 |
+
json_path = os.path.join(self.data_dir, 'jsons', f"{file_name}.json")
|
61 |
+
img_dir = 'images'
|
62 |
+
elif file_name in os.listdir(os.path.join(self.data_dir, 'jsons2')):
|
63 |
+
json_path = os.path.join(self.data_dir, 'jsons2', f"{file_name}.json")
|
64 |
+
img_dir = 'images2'
|
65 |
+
else:
|
66 |
+
json_path = os.path.join(self.data_dir, 'jsons3', f"{file_name}.json")
|
67 |
+
img_dir = 'images3'
|
68 |
+
|
69 |
+
# Load questions from the JSON file
|
70 |
with open(json_path, 'r') as f:
|
71 |
json_data = json.load(f)
|
72 |
for item in json_data:
|
73 |
if 'PMC_ID' not in item or 'qa_id' not in item:
|
74 |
continue # Ensure all necessary fields are present
|
75 |
+
item['image_path'] = os.path.join(self.data_dir, img_dir, f"{item['PMC_ID']}.jpg")
|
76 |
if os.path.exists(item['image_path']):
|
77 |
self.questions.append(item)
|
78 |
else:
|
79 |
unused_count += 1
|
80 |
+
print(f"Total unused/used images: {unused_count} / {len(self.file_names) - unused_count}")
|
|
|
|
|
|
|
81 |
|
82 |
def __len__(self):
|
83 |
return len(self.questions)
|
84 |
|
85 |
def __getitem__(self, idx):
|
|
|
|
|
|
|
86 |
"""
|
87 |
Loads a single data point.
|
88 |
|
|
|
134 |
}
|
135 |
|
136 |
if __name__ == "__main__":
|
137 |
+
# Initialize dataset for training
|
138 |
+
dataset = RQADataset(data_dir='.', split='train')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
139 |
|
140 |
# Test loading a single item
|
141 |
print(f"Number of samples in dataset: {len(dataset)}")
|