File size: 11,089 Bytes
fdeced5
 
 
 
32e5fcf
fdeced5
 
32e5fcf
 
 
 
 
 
 
 
fdeced5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32e5fcf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
fdeced5
 
 
 
 
2e88e30
 
 
 
 
 
 
 
fdeced5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
32e5fcf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
"""Input loader stage for the character attribute extraction pipeline."""

import os
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple, Union, Iterator
from PIL import Image
import logging
import torch
from torch.utils.data import Dataset, DataLoader
try:
    from datasets import Dataset as HFDataset, load_dataset
    HF_AVAILABLE = True
except ImportError:
    HF_AVAILABLE = False
    HFDataset = None

from .base import PipelineStage

logger = logging.getLogger(__name__)

class DatasetItem:
    """Represents a single item from the dataset."""
    
    def __init__(self, image_path: str, text_path: Optional[str] = None, tags: Optional[str] = None):
        self.image_path = image_path
        self.text_path = text_path
        self.tags = tags
        self.item_id = Path(image_path).stem
    
    def load_image(self) -> Image.Image:
        """Load and return the PIL Image."""
        try:
            return Image.open(self.image_path).convert('RGB')
        except Exception as e:
            logger.error(f"Failed to load image {self.image_path}: {e}")
            raise
    
    def load_tags(self) -> str:
        """Load tags from text file or return provided tags."""
        if self.tags:
            return self.tags
        
        if self.text_path and os.path.exists(self.text_path):
            try:
                with open(self.text_path, 'r', encoding='utf-8') as f:
                    return f.read().strip()
            except Exception as e:
                logger.error(f"Failed to load tags from {self.text_path}: {e}")
                return ""
        
        return ""

class CharacterDataset(Dataset):
    """PyTorch Dataset for character images and attributes."""
    
    def __init__(self, items: List[DatasetItem], transform=None):
        self.items = items
        self.transform = transform
    
    def __len__(self):
        return len(self.items)
    
    def __getitem__(self, idx):
        item = self.items[idx]
        image = item.load_image()
        tags = item.load_tags()
        
        if self.transform:
            image = self.transform(image)
        
        return {
            'image': image,
            'tags': tags,
            'item_id': item.item_id,
            'image_path': item.image_path
        }

class InputLoader(PipelineStage):
    """Loads images and associated text data from the dataset."""
    
    def __init__(self, config: Optional[Dict[str, Any]] = None):
        super().__init__("InputLoader", config)
        if config:
            self.dataset_path = config.get('dataset_path', './continued/sensitive')
            self.batch_size = config.get('batch_size', 32)
            self.num_workers = config.get('num_workers', 4)
        else:
            self.dataset_path = './continued/sensitive'
            self.batch_size = 32
            self.num_workers = 4
        self.supported_image_formats = {'.jpg', '.jpeg', '.png', '.bmp', '.tiff'}
    
    def discover_dataset_items(self) -> List[DatasetItem]:
        """Discover all image-text pairs in the dataset directory."""
        dataset_path = Path(self.dataset_path)
        items = []
        
        if not dataset_path.exists():
            self.logger.error(f"Dataset path does not exist: {dataset_path}")
            return items
        
        # Find all image files
        image_files = []
        for ext in self.supported_image_formats:
            image_files.extend(dataset_path.glob(f"*{ext}"))
        
        self.logger.info(f"Found {len(image_files)} image files")
        
        for image_path in image_files:
            # Look for corresponding text file
            text_path = image_path.with_suffix('.txt')
            
            item = DatasetItem(
                image_path=str(image_path),
                text_path=str(text_path) if text_path.exists() else None
            )
            items.append(item)
        
        self.logger.info(f"Created {len(items)} dataset items")
        return items
    
    def load_single_item(self, item_path: Union[str, Path]) -> DatasetItem:
        """Load a single item by path."""
        item_path = Path(item_path)
        
        if not item_path.exists():
            raise FileNotFoundError(f"Item not found: {item_path}")
        
        # If it's an image, look for corresponding text
        if item_path.suffix.lower() in self.supported_image_formats:
            text_path = item_path.with_suffix('.txt')
            return DatasetItem(
                image_path=str(item_path),
                text_path=str(text_path) if text_path.exists() else None
            )
        
        # If it's a text file, look for corresponding image
        elif item_path.suffix.lower() == '.txt':
            for ext in self.supported_image_formats:
                image_path = item_path.with_suffix(ext)
                if image_path.exists():
                    return DatasetItem(
                        image_path=str(image_path),
                        text_path=str(item_path)
                    )
            
            # No corresponding image found, create text-only item
            with open(item_path, 'r', encoding='utf-8') as f:
                tags = f.read().strip()
            
            return DatasetItem(
                image_path=None,
                text_path=str(item_path),
                tags=tags
            )
        
        else:
            raise ValueError(f"Unsupported file format: {item_path.suffix}")
    
    def process(self, input_data: Any) -> Dict[str, Any]:
        """Process input data and return loaded content."""
        if isinstance(input_data, (str, Path)):
            # Single item path provided
            item = self.load_single_item(input_data)
        elif isinstance(input_data, DatasetItem):
            # DatasetItem provided directly
            item = input_data
        else:
            raise ValueError(f"Unsupported input type: {type(input_data)}")
        
        result = {
            'item_id': item.item_id,
            'image': None,
            'tags': '',
            'image_path': item.image_path,
            'text_path': item.text_path
        }
        
        # Load image if available
        if item.image_path and os.path.exists(item.image_path):
            try:
                result['image'] = item.load_image()
                self.logger.debug(f"Loaded image: {item.image_path}")
            except Exception as e:
                self.logger.error(f"Failed to load image {item.image_path}: {e}")
                result['image'] = None
        
        # Load tags if available
        try:
            result['tags'] = item.load_tags()
            self.logger.debug(f"Loaded tags: {len(result['tags'])} characters")
        except Exception as e:
            self.logger.error(f"Failed to load tags: {e}")
            result['tags'] = ''
        
        return result
    
    def validate_input(self, input_data: Any) -> bool:
        """Validate that input can be processed."""
        if isinstance(input_data, (str, Path)):
            return Path(input_data).exists()
        elif isinstance(input_data, DatasetItem):
            return True
        return False
    
    def get_sample_items(self, n: int = 10) -> List[DatasetItem]:
        """Get a sample of dataset items for testing."""
        all_items = self.discover_dataset_items()
        return all_items[:n] if len(all_items) >= n else all_items
    
    def create_pytorch_dataset(self, items: Optional[List[DatasetItem]] = None, transform=None) -> CharacterDataset:
        """Create a PyTorch Dataset from dataset items."""
        if items is None:
            items = self.discover_dataset_items()
        return CharacterDataset(items, transform=transform)
    
    def create_dataloader(self, items: Optional[List[DatasetItem]] = None, transform=None, 
                         batch_size: Optional[int] = None, shuffle: bool = True) -> DataLoader:
        """Create a PyTorch DataLoader for batch processing."""
        dataset = self.create_pytorch_dataset(items, transform)
        batch_size = batch_size or self.batch_size
        
        return DataLoader(
            dataset,
            batch_size=batch_size,
            shuffle=shuffle,
            num_workers=self.num_workers,
            collate_fn=self._collate_fn
        )
    
    def _collate_fn(self, batch):
        """Custom collate function for DataLoader."""
        images = [item['image'] for item in batch]
        tags = [item['tags'] for item in batch]
        item_ids = [item['item_id'] for item in batch]
        image_paths = [item['image_path'] for item in batch]
        
        return {
            'images': images,
            'tags': tags,
            'item_ids': item_ids,
            'image_paths': image_paths
        }
    
    def create_huggingface_dataset(self, items: Optional[List[DatasetItem]] = None) -> Optional[HFDataset]:
        """Create a HuggingFace Dataset for efficient batch processing."""
        if not HF_AVAILABLE:
            logger.warning("HuggingFace datasets not available. Install with: pip install datasets")
            return None
        
        if items is None:
            items = self.discover_dataset_items()
        
        # Prepare data for HuggingFace Dataset
        data = {
            'image_path': [item.image_path for item in items],
            'tags': [item.load_tags() for item in items],
            'item_id': [item.item_id for item in items]
        }
        
        return HFDataset.from_dict(data)
    
    def process_with_hf_map(self, processing_fn, items: Optional[List[DatasetItem]] = None, 
                           batch_size: Optional[int] = None, num_proc: int = 4) -> Optional[HFDataset]:
        """Process dataset using HuggingFace datasets.map() for efficient batch inference."""
        if not HF_AVAILABLE:
            logger.warning("HuggingFace datasets not available")
            return None
        
        dataset = self.create_huggingface_dataset(items)
        if dataset is None:
            return None
        
        batch_size = batch_size or self.batch_size
        
        # Apply processing function using datasets.map()
        processed_dataset = dataset.map(
            processing_fn,
            batched=True,
            batch_size=batch_size,
            num_proc=num_proc,
            remove_columns=['image_path'],  # Remove to save memory
            desc="Processing character attributes"
        )
        
        return processed_dataset
    
    def batch_process_iterator(self, items: Optional[List[DatasetItem]] = None, 
                              batch_size: Optional[int] = None) -> Iterator[List[DatasetItem]]:
        """Create an iterator for batch processing without loading all data into memory."""
        if items is None:
            items = self.discover_dataset_items()
        
        batch_size = batch_size or self.batch_size
        
        for i in range(0, len(items), batch_size):
            yield items[i:i + batch_size]