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

Discard previous results for filtering if the engine is changed

Browse Source
This commit is contained in:
AuroraWright
2024-06-25 09:26:44 +02:00
parent 1d5206db5b
commit d0622f4b83
1 changed files with 14 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
owocr/run.py
View File

@@ -269,7 +269,10 @@ class TextFiltering:
def __call__(self, text, last_text): def __call__(self, text, last_text):
orig_text = self.segmenter.segment(text) orig_text = self.segmenter.segment(text)
new_blocks = [block for block in orig_text if block not in last_text] if last_text[1] != engine_index:
new_blocks = orig_text
else:
new_blocks = [block for block in orig_text if block not in last_text[0]]
final_blocks = [] final_blocks = []
if self.accurate_filtering: if self.accurate_filtering:
detection_results = self.pipe(new_blocks, top_k=2, truncation=True) detection_results = self.pipe(new_blocks, top_k=2, truncation=True)
@@ -453,7 +456,7 @@ def are_images_identical(img1, img2):
return (img1.shape == img2.shape) and (img1 == img2).all() return (img1.shape == img2.shape) and (img1 == img2).all()
def process_and_write_results(img_or_path, write_to, notifications, enable_filtering, last_text, filtering): def process_and_write_results(img_or_path, write_to, notifications, last_text, filtering):
engine_instance = engine_instances[engine_index] engine_instance = engine_instances[engine_index]
t0 = time.time() t0 = time.time()
res, text = engine_instance(img_or_path) res, text = engine_instance(img_or_path)
@@ -462,7 +465,7 @@ def process_and_write_results(img_or_path, write_to, notifications, enable_filte
orig_text = '' orig_text = ''
engine_color = config.get_general('engine_color') engine_color = config.get_general('engine_color')
if res: if res:
if enable_filtering: if filtering:
text, orig_text = filtering(text, last_text) text, orig_text = filtering(text, last_text)
text = post_process(text) text = post_process(text)
logger.opt(ansi=True).info(f'Text recognized in {t1 - t0:0.03f}s using <{engine_color}>{engine_instance.readable_name}</{engine_color}>: {text}') logger.opt(ansi=True).info(f'Text recognized in {t1 - t0:0.03f}s using <{engine_color}>{engine_instance.readable_name}</{engine_color}>: {text}')
@@ -479,7 +482,7 @@ def process_and_write_results(img_or_path, write_to, notifications, enable_filte
else: else:
logger.opt(ansi=True).info(f'<{engine_color}>{engine_instance.readable_name}</{engine_color}> reported an error after {t1 - t0:0.03f}s: {text}') logger.opt(ansi=True).info(f'<{engine_color}>{engine_instance.readable_name}</{engine_color}> reported an error after {t1 - t0:0.03f}s: {text}')
return orig_text return (orig_text, engine_index)
def get_path_key(path): def get_path_key(path):
@@ -654,7 +657,7 @@ def run(read_from=None,
screencapture_mode = None screencapture_mode = None
screencapture_window_active = True screencapture_window_active = True
screencapture_window_visible = True screencapture_window_visible = True
last_text = [] last_text = ([], engine_index)
if screen_capture_coords == '': if screen_capture_coords == '':
screencapture_mode = 0 screencapture_mode = 0
elif len(screen_capture_coords.split(',')) == 4: elif len(screen_capture_coords.split(',')) == 4:
@@ -792,7 +795,7 @@ def run(read_from=None,
else: else:
if not paused: if not paused:
img = Image.open(io.BytesIO(item)) img = Image.open(io.BytesIO(item))
process_and_write_results(img, write_to, notifications, False, '', None) process_and_write_results(img, write_to, notifications, '', None)
elif read_from == 'unixsocket': elif read_from == 'unixsocket':
while True: while True:
try: try:
@@ -802,7 +805,7 @@ def run(read_from=None,
else: else:
if not paused: if not paused:
img = Image.open(io.BytesIO(item)) img = Image.open(io.BytesIO(item))
process_and_write_results(img, write_to, notifications, False, '', None) process_and_write_results(img, write_to, notifications, '', None)
elif read_from == 'clipboard': elif read_from == 'clipboard':
process_clipboard = False process_clipboard = False
if windows_clipboard_polling: if windows_clipboard_polling:
@@ -850,7 +853,7 @@ def run(read_from=None,
process_clipboard = True process_clipboard = True
if process_clipboard: if process_clipboard:
process_and_write_results(img, write_to, notifications, False, '', None) process_and_write_results(img, write_to, notifications, '', None)
just_unpaused = False just_unpaused = False
@@ -906,8 +909,8 @@ def run(read_from=None,
else: else:
sct_img = sct.grab(sct_params) sct_img = sct.grab(sct_params)
img = Image.frombytes('RGB', sct_img.size, sct_img.bgra, 'raw', 'BGRX') img = Image.frombytes('RGB', sct_img.size, sct_img.bgra, 'raw', 'BGRX')
res = process_and_write_results(img, write_to, notifications, True, last_text, filtering) res = process_and_write_results(img, write_to, notifications, last_text, filtering)
if res != '': if res[0] != '':
last_text = res last_text = res
delay = screen_capture_delay_secs delay = screen_capture_delay_secs
else: else:
@@ -929,7 +932,7 @@ def run(read_from=None,
except (UnidentifiedImageError, OSError) as e: except (UnidentifiedImageError, OSError) as e:
logger.warning(f'Error while reading file {path}: {e}') logger.warning(f'Error while reading file {path}: {e}')
else: else:
process_and_write_results(img, write_to, notifications, False, '', None) process_and_write_results(img, write_to, notifications, '', None)
img.close() img.close()
if delete_images: if delete_images:
Path.unlink(path) Path.unlink(path)