Commit
·
ef2032a
1
Parent(s):
84b78a0
printing extracted text22
Browse files- HTR/app.py +11 -2
- all_models.py +41 -24
- main.py +7 -1
HTR/app.py
CHANGED
@@ -17,8 +17,17 @@ from HTR.strike import struck_images
|
|
17 |
from HTR.hcr import text
|
18 |
from HTR.spell_and_gramer_check import spell_grammer
|
19 |
|
20 |
-
#
|
21 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
|
23 |
def preprocess_image(img):
|
24 |
"""Preprocess image to improve text detection"""
|
|
|
17 |
from HTR.hcr import text
|
18 |
from HTR.spell_and_gramer_check import spell_grammer
|
19 |
|
20 |
+
# Get absolute path to project root
|
21 |
+
project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
22 |
+
model_path = os.path.join(project_root, 'models', 'vit-base-beans')
|
23 |
+
|
24 |
+
# Log model path for debugging
|
25 |
+
logger.info(f"Using model path: {model_path}")
|
26 |
+
if not os.path.exists(model_path):
|
27 |
+
logger.error(f"Model directory not found at: {model_path}")
|
28 |
+
else:
|
29 |
+
files = os.listdir(model_path)
|
30 |
+
logger.info(f"Found model files: {files}")
|
31 |
|
32 |
def preprocess_image(img):
|
33 |
"""Preprocess image to improve text detection"""
|
all_models.py
CHANGED
@@ -271,52 +271,69 @@ class ModelSingleton:
|
|
271 |
raise
|
272 |
|
273 |
def get_vit_model(self):
|
274 |
-
"""Get ViT model
|
275 |
try:
|
276 |
if self.vit_model is None:
|
277 |
-
from transformers import ViTImageProcessor, ViTForImageClassification
|
278 |
-
logger.info("Loading local ViT model...")
|
279 |
|
280 |
-
#
|
281 |
-
|
282 |
-
|
|
|
283 |
|
|
|
284 |
if not os.path.exists(model_path):
|
285 |
raise FileNotFoundError(f"Local model directory not found at: {model_path}")
|
286 |
|
287 |
-
#
|
288 |
-
|
289 |
-
|
290 |
-
if missing_files:
|
291 |
-
raise FileNotFoundError(f"Missing required model files: {missing_files}")
|
292 |
|
293 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
294 |
self.vit_processor = ViTImageProcessor(
|
295 |
do_resize=True,
|
296 |
-
size=
|
297 |
-
do_normalize=True
|
298 |
-
image_mean=[0.5, 0.5, 0.5],
|
299 |
-
image_std=[0.5, 0.5, 0.5]
|
300 |
)
|
301 |
|
302 |
-
# Load model from local files
|
303 |
-
logger.info("Loading
|
304 |
self.vit_model = ViTForImageClassification.from_pretrained(
|
305 |
model_path,
|
306 |
-
|
307 |
-
|
308 |
-
|
|
|
|
|
|
|
309 |
)
|
310 |
|
311 |
-
|
312 |
self.vit_model.to(self.device)
|
313 |
self.vit_model.eval()
|
314 |
-
logger.info(
|
315 |
|
316 |
self._reference_counts['vit'] += 1
|
317 |
return self.vit_model, self.vit_processor
|
|
|
318 |
except Exception as e:
|
319 |
-
logger.error(f"Error loading local ViT model: {e}")
|
320 |
raise
|
321 |
|
322 |
def release_similarity_model(self):
|
|
|
271 |
raise
|
272 |
|
273 |
def get_vit_model(self):
|
274 |
+
"""Get ViT model using only local files - no downloads"""
|
275 |
try:
|
276 |
if self.vit_model is None:
|
277 |
+
from transformers import ViTConfig, ViTImageProcessor, ViTForImageClassification
|
278 |
+
logger.info("Loading local ViT model from files...")
|
279 |
|
280 |
+
# Get absolute path to model directory
|
281 |
+
model_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
|
282 |
+
model_path = os.path.join(model_root, 'models', 'vit-base-beans')
|
283 |
+
logger.info(f"Using local model directory: {model_path}")
|
284 |
|
285 |
+
# Check model directory exists
|
286 |
if not os.path.exists(model_path):
|
287 |
raise FileNotFoundError(f"Local model directory not found at: {model_path}")
|
288 |
|
289 |
+
# Get paths to required files
|
290 |
+
model_file = os.path.join(model_path, 'model.safetensors')
|
291 |
+
config_file = os.path.join(model_path, 'config.json')
|
|
|
|
|
292 |
|
293 |
+
# Verify files exist
|
294 |
+
if not os.path.exists(model_file):
|
295 |
+
raise FileNotFoundError(f"Local model weights file not found at: {model_file}")
|
296 |
+
if not os.path.exists(config_file):
|
297 |
+
raise FileNotFoundError(f"Local model config file not found at: {config_file}")
|
298 |
+
|
299 |
+
logger.info("Found all required local model files:")
|
300 |
+
logger.info(f"- Using model weights: {model_file}")
|
301 |
+
logger.info(f"- Using config file: {config_file}")
|
302 |
+
|
303 |
+
# Load config directly from file
|
304 |
+
logger.info("Loading model configuration from local file...")
|
305 |
+
config = ViTConfig.from_json_file(config_file)
|
306 |
+
|
307 |
+
# Create processor from local config
|
308 |
+
logger.info("Creating image processor from local config...")
|
309 |
self.vit_processor = ViTImageProcessor(
|
310 |
do_resize=True,
|
311 |
+
size=config.image_size,
|
312 |
+
do_normalize=True
|
|
|
|
|
313 |
)
|
314 |
|
315 |
+
# Load model directly from local files
|
316 |
+
logger.info("Loading model weights from local file...")
|
317 |
self.vit_model = ViTForImageClassification.from_pretrained(
|
318 |
model_path,
|
319 |
+
config=config,
|
320 |
+
local_files_only=True,
|
321 |
+
use_safetensors=True,
|
322 |
+
trust_remote_code=False,
|
323 |
+
from_tf=False,
|
324 |
+
_fast_init=True
|
325 |
)
|
326 |
|
327 |
+
logger.info(f"Moving model to {self.device}...")
|
328 |
self.vit_model.to(self.device)
|
329 |
self.vit_model.eval()
|
330 |
+
logger.info("Local model loaded successfully!")
|
331 |
|
332 |
self._reference_counts['vit'] += 1
|
333 |
return self.vit_model, self.vit_processor
|
334 |
+
|
335 |
except Exception as e:
|
336 |
+
logger.error(f"Error loading local ViT model: {str(e)}")
|
337 |
raise
|
338 |
|
339 |
def release_similarity_model(self):
|
main.py
CHANGED
@@ -52,6 +52,11 @@ def get_user_cache_dir():
|
|
52 |
BASE_DIR = get_user_cache_dir()
|
53 |
log_print(f"Using base directory: {BASE_DIR}")
|
54 |
|
|
|
|
|
|
|
|
|
|
|
55 |
# Set environment variables before any other imports
|
56 |
cache_dirs = {
|
57 |
'root': BASE_DIR,
|
@@ -65,7 +70,8 @@ cache_dirs = {
|
|
65 |
'logs': os.path.join(BASE_DIR, 'logs'),
|
66 |
'uploads': os.path.join(BASE_DIR, 'uploads'),
|
67 |
'images': os.path.join(BASE_DIR, 'images'),
|
68 |
-
'ans_image': os.path.join(BASE_DIR, 'ans_image')
|
|
|
69 |
}
|
70 |
|
71 |
# Create all necessary directories with proper permissions
|
|
|
52 |
BASE_DIR = get_user_cache_dir()
|
53 |
log_print(f"Using base directory: {BASE_DIR}")
|
54 |
|
55 |
+
# Get absolute path to project root for models
|
56 |
+
PROJECT_ROOT = os.path.dirname(os.path.abspath(__file__))
|
57 |
+
os.environ['MODEL_ROOT'] = PROJECT_ROOT
|
58 |
+
log_print(f"Set MODEL_ROOT to: {PROJECT_ROOT}")
|
59 |
+
|
60 |
# Set environment variables before any other imports
|
61 |
cache_dirs = {
|
62 |
'root': BASE_DIR,
|
|
|
70 |
'logs': os.path.join(BASE_DIR, 'logs'),
|
71 |
'uploads': os.path.join(BASE_DIR, 'uploads'),
|
72 |
'images': os.path.join(BASE_DIR, 'images'),
|
73 |
+
'ans_image': os.path.join(BASE_DIR, 'ans_image'),
|
74 |
+
'models': os.path.join(PROJECT_ROOT, 'models') # Add models directory
|
75 |
}
|
76 |
|
77 |
# Create all necessary directories with proper permissions
|