sudacode/owocr
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix bounding box issues with resized images

Browse Source
This commit is contained in:
AuroraWright
2025-10-14 15:14:21 +02:00
parent 3dc55c4702
commit a86b0c2dc9
2 changed files with 55 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
owocr/ocr.py
View File

@@ -170,7 +170,7 @@ def pil_image_to_numpy_array(img):
def limit_image_size(img, max_size): def limit_image_size(img, max_size):
img_bytes = pil_image_to_bytes(img) img_bytes = pil_image_to_bytes(img)
if len(img_bytes) <= max_size: if len(img_bytes) <= max_size:
return img_bytes, 'png' return img_bytes, 'png', img.size
scaling_factor = 0.60 if any(x > 2000 for x in img.size) else 0.75 scaling_factor = 0.60 if any(x > 2000 for x in img.size) else 0.75
new_w = int(img.width * scaling_factor) new_w = int(img.width * scaling_factor)
@@ -178,18 +178,18 @@ def limit_image_size(img, max_size):
resized_img = img.resize((new_w, new_h), Image.Resampling.LANCZOS) resized_img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
resized_img_bytes = pil_image_to_bytes(resized_img) resized_img_bytes = pil_image_to_bytes(resized_img)
if len(resized_img_bytes) <= max_size: if len(resized_img_bytes) <= max_size:
return resized_img_bytes, 'png' return resized_img_bytes, 'png', resized_img.size
for _ in range(2): for _ in range(2):
jpeg_quality = 80 jpeg_quality = 80
while jpeg_quality >= 60: while jpeg_quality >= 60:
img_bytes = pil_image_to_bytes(img, 'jpeg', jpeg_quality=jpeg_quality, optimize=True) img_bytes = pil_image_to_bytes(img, 'jpeg', jpeg_quality=jpeg_quality, optimize=True)
if len(img_bytes) <= max_size: if len(img_bytes) <= max_size:
return img_bytes, 'jpeg' return img_bytes, 'jpeg', img.size
jpeg_quality -= 5 jpeg_quality -= 5
img = resized_img img = resized_img
return False, '' return False, '', (None, None)
class MangaOcr: class MangaOcr:
@@ -603,7 +603,7 @@ class Bing:
if not img: if not img:
return (False, 'Invalid image provided') return (False, 'Invalid image provided')
img_bytes = self._preprocess(img) img_bytes, img_size = self._preprocess(img)
if not img_bytes: if not img_bytes:
return (False, 'Image is too big!') return (False, 'Image is too big!')
@@ -676,7 +676,8 @@ class Bing:
data = res.json() data = res.json()
ocr_result = self._to_generic_result(data, img.width, img.height) img_width, img_height = img_size
ocr_result = self._to_generic_result(data, img_width, img_height)
x = (True, ocr_result) x = (True, ocr_result)
if is_path: if is_path:
@@ -694,12 +695,12 @@ class Bing:
new_h = int(img.height * resize_factor) new_h = int(img.height * resize_factor)
img = img.resize((new_w, new_h), Image.Resampling.LANCZOS) img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
img_bytes, _ = limit_image_size(img, max_byte_size) img_bytes, _, img_size = limit_image_size(img, max_byte_size)
if img_bytes: if img_bytes:
res = base64.b64encode(img_bytes).decode('utf-8') res = base64.b64encode(img_bytes).decode('utf-8')
return res return res, img_size
class AppleVision: class AppleVision:
name = 'avision' name = 'avision'
@@ -1053,13 +1054,16 @@ class OneOCR:
return (False, 'Invalid image provided') return (False, 'Invalid image provided')
if sys.platform == 'win32': if sys.platform == 'win32':
img_processed = self._preprocess_windows(img)
img_width, img_height = img_processed.size
try: try:
raw_res = self.model.recognize_pil(img) raw_res = self.model.recognize_pil(img_processed)
except RuntimeError as e: except RuntimeError as e:
return (False, e) return (False, e)
else: else:
img_processed, img_width, img_height = self._preprocess_notwindows(img)
try: try:
res = requests.post(self.url, data=self._preprocess(img), timeout=3) res = requests.post(self.url, data=img_processed, timeout=3)
except requests.exceptions.Timeout: except requests.exceptions.Timeout:
return (False, 'Request timeout!') return (False, 'Request timeout!')
except requests.exceptions.ConnectionError: except requests.exceptions.ConnectionError:
@@ -1070,15 +1074,37 @@ class OneOCR:
raw_res = res.json() raw_res = res.json()
ocr_response = self._to_generic_result(raw_res, img.width, img.height) if 'error' in raw_res:
return (False, raw_res['error'])
ocr_response = self._to_generic_result(raw_res, img_width, img_height)
x = (True, ocr_response) x = (True, ocr_response)
if is_path: if is_path:
img.close() img.close()
return x return x
def _preprocess(self, img): def _preprocess_windows(self, img):
return pil_image_to_bytes(img, png_compression=1) min_pixel_size = 50
max_pixel_size = 10000
if any(x < min_pixel_size for x in img.size):
resize_factor = max(min_pixel_size / img.width, min_pixel_size / img.height)
new_w = int(img.width * resize_factor)
new_h = int(img.height * resize_factor)
img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
if any(x > max_pixel_size for x in img.size):
resize_factor = min(max_pixel_size / img.width, max_pixel_size / img.height)
new_w = int(img.width * resize_factor)
new_h = int(img.height * resize_factor)
img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
return img
def _preprocess_notwindows(self, img):
img = self._preprocess_windows(img)
return pil_image_to_bytes(img, png_compression=1), img.width, img.height
class AzureImageAnalysis: class AzureImageAnalysis:
name = 'azure' name = 'azure'
@@ -1129,8 +1155,17 @@ class AzureImageAnalysis:
return x return x
def _preprocess(self, img): def _preprocess(self, img):
if any(x < 50 for x in img.size): min_pixel_size = 50
resize_factor = max(50 / img.width, 50 / img.height) max_pixel_size = 10000
if any(x < min_pixel_size for x in img.size):
resize_factor = max(min_pixel_size / img.width, min_pixel_size / img.height)
new_w = int(img.width * resize_factor)
new_h = int(img.height * resize_factor)
img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
if any(x > max_pixel_size for x in img.size):
resize_factor = min(max_pixel_size / img.width, max_pixel_size / img.height)
new_w = int(img.width * resize_factor) new_w = int(img.width * resize_factor)
new_h = int(img.height * resize_factor) new_h = int(img.height * resize_factor)
img = img.resize((new_w, new_h), Image.Resampling.LANCZOS) img = img.resize((new_w, new_h), Image.Resampling.LANCZOS)
@@ -1286,7 +1321,7 @@ class OCRSpace:
if not img: if not img:
return (False, 'Invalid image provided') return (False, 'Invalid image provided')
img_bytes, img_extension = self._preprocess(img) img_bytes, img_extension, _ = self._preprocess(img)
if not img_bytes: if not img_bytes:
return (False, 'Image is too big!') return (False, 'Image is too big!')

12
owocr/run.py
View File

@@ -1526,8 +1526,8 @@ def run():
engine_secondary = engine_class.key engine_secondary = engine_class.key
if len(engine_keys) == 0: if len(engine_keys) == 0:
msg = 'No engines available!' logger.error('No engines available!')
raise NotImplementedError(msg) sys.exit(1)
global engine_index global engine_index
global engine_index_2 global engine_index_2
@@ -1545,12 +1545,12 @@ def run():
read_from_path = None read_from_path = None
read_from_readable = [] read_from_readable = []
write_to = config.get_general('write_to') write_to = config.get_general('write_to')
output_format = config.get_general('output_format')
terminated = threading.Event() terminated = threading.Event()
paused = threading.Event() paused = threading.Event()
if config.get_general('pause_at_startup'): if config.get_general('pause_at_startup'):
paused.set() paused.set()
auto_pause = config.get_general('auto_pause') auto_pause = config.get_general('auto_pause')
output_format = config.get_general('output_format')
clipboard_thread = None clipboard_thread = None
websocket_server_thread = None websocket_server_thread = None
screenshot_thread = None screenshot_thread = None
@@ -1651,12 +1651,6 @@ def run():
user_input_thread = threading.Thread(target=user_input_thread_run, daemon=True) user_input_thread = threading.Thread(target=user_input_thread_run, daemon=True)
user_input_thread.start() user_input_thread.start()
if output_format == 'json' and not engine_instances[engine_index].coordinate_support:
supported_engines = (engine.name for engine in engine_instances if engine.coordinate_support)
logger.error(f"The selected engine '{engine_instances[engine_index].name}' does not support coordinate output.")
logger.error(f"Please choose one of: {', '.join(supported_engines)}.")
sys.exit(1)
logger.opt(colors=True).info(f"Reading from {' and '.join(read_from_readable)}, writing to {write_to_readable} using <{engine_color}>{engine_instances[engine_index].readable_name}</{engine_color}>{' (paused)' if paused.is_set() else ''}") logger.opt(colors=True).info(f"Reading from {' and '.join(read_from_readable)}, writing to {write_to_readable} using <{engine_color}>{engine_instances[engine_index].readable_name}</{engine_color}>{' (paused)' if paused.is_set() else ''}")
while not terminated.is_set(): while not terminated.is_set():