diff --git a/README.md b/README.md
index e79429e..dc188ac 100644
--- a/README.md
+++ b/README.md
@@ -61,6 +61,7 @@ The command-line options/config file allow you to configure OCR providers, hotke
- Apple Live Text (VisionKit framework) - It should be the same as Vision except that in Sonoma Apple added vertical text reading. **macOS only - Recommended (pre-installed)** → key: `d`
- WinRT OCR: install: `pip install "owocr[winocr]"`. It can also be used by installing winocr on a Windows virtual machine and running the server there (`winocr_serve`) and specifying the IP address of the Windows VM/machine in the config file. **Windows 10/11 only** → key: `w`
- OneOCR - install: `pip install "owocr[oneocr]"`. Close second local best to the Apple one. You need to copy 3 system files from Windows 11 to use it, refer to the readme [here](https://github.com/AuroraWright/oneocr). It can also be used by installing oneocr on a Windows virtual machine and running the server there (`oneocr_serve`) and specifying the IP address of the Windows VM/machine in the config file. **Windows 10/11 only - Recommended** → key: `z`
+- [meikiocr](https://github.com/rtr46/meikiocr) - install: `pip install "owocr[meikiocr]"`. Comparable to OneOCR in accuracy and CPU latency. Can be run on Nvidia GPUs via `pip uninstall onnxruntime && pip install onnxruntime-gpu` making it the fastest OCR available. Probably best option for Linux users. Can't process vertical text and is limited to 64 text lines and 48 characters per line. → key: `k`
## Cloud
- Google Lens - install: `pip install "owocr[lens]"`. Arguably the best OCR engine to date. **Recommended** → key: `l`
diff --git a/owocr/ocr.py b/owocr/ocr.py
index bef28f8..401cb06 100644
--- a/owocr/ocr.py
+++ b/owocr/ocr.py
@@ -62,6 +62,11 @@ try:
except ImportError:
pass
+try:
+ from meikiocr import MeikiOCR as MKOCR
+except ImportError:
+ pass
+
try:
import winocr
except ImportError:
@@ -1775,6 +1780,104 @@ class RapidOCR:
def _preprocess(self, img):
return pil_image_to_numpy_array(img)
+class MeikiOCR:
+ name = 'meikiocr'
+ readable_name = 'meikiocr'
+ key = 'k'
+ config_entry = 'meikiocr'
+ available = False
+ local = True
+ manual_language = False
+ coordinate_support = True
+ threading_support = True
+
+ def __init__(self, config={}, language='ja'):
+ if 'meikiocr' not in sys.modules:
+ logger.warning('meikiocr not available, meikiocr will not work!')
+ else:
+ logger.info('Loading meikiocr model')
+ self.model = MKOCR()
+ self.available = True
+ logger.info('meikiocr ready')
+
+ def _to_normalized_bbox(self, pixel_bbox: List[int], img_width: int, img_height: int) -> BoundingBox:
+ """Converts a bbox [x1, y1, x2, y2] to a normalized BoundingBox."""
+ x1, y1, x2, y2 = pixel_bbox
+ width_px = x2 - x1
+ height_px = y2 - y1
+ center_x_px = x1 + width_px / 2
+ center_y_px = y1 + height_px / 2
+
+ return BoundingBox(
+ center_x=center_x_px / img_width,
+ center_y=center_y_px / img_height,
+ width=width_px / img_width,
+ height=height_px / img_height
+ )
+
+ def _to_generic_result(self, response: List[dict], img_width: int, img_height: int) -> OcrResult:
+ """Converts the raw meikiocr output into the standardized OcrResult format."""
+ paragraphs = []
+
+ # each dictionary in the response corresponds to a detected line of text.
+ # treat each line as a separate Paragraph containing a single Line.
+ for line_result in response:
+ line_text = line_result.get('text', '')
+ char_results = line_result.get('chars', [])
+ if not line_text or not char_results:
+ continue
+
+ char_in_line = []
+ for char_info in char_results:
+ normalized_bbox = self._to_normalized_bbox(
+ char_info['bbox'], img_width, img_height
+ )
+ word = Word(
+ text=char_info['char'],
+ bounding_box=normalized_bbox
+ )
+ char_in_line.append(word)
+
+ if not char_in_line:
+ continue
+
+ line_bbox = merge_bounding_boxes(char_in_line)
+
+ line = Line(
+ bounding_box=line_bbox,
+ words=char_in_line,
+ text=line_text
+ )
+
+ # each line becomes a paragraph.
+ paragraph = Paragraph(
+ bounding_box=line_bbox,
+ lines=[line],
+ writing_direction="LEFT_TO_RIGHT" # meikiocr only supports horizontal text
+ )
+ paragraphs.append(paragraph)
+
+ return OcrResult(
+ image_properties=ImageProperties(width=img_width, height=img_height),
+ paragraphs=paragraphs
+ )
+
+ def __call__(self, img):
+ img, is_path = input_to_pil_image(img)
+ if not img:
+ return (False, 'Invalid image provided')
+
+ image_np = np.array(img.convert('RGB'))[:, :, ::-1]
+
+ read_results = self.model.run_ocr(image_np)
+ ocr_result = self._to_generic_result(read_results, img.width, img.height)
+
+ x = (True, ocr_result)
+
+ if is_path:
+ img.close()
+ return x
+
class OCRSpace:
name = 'ocrspace'
readable_name = 'OCRSpace'
diff --git a/owocr_config.ini b/owocr_config.ini
index b63335c..de31c1c 100644
--- a/owocr_config.ini
+++ b/owocr_config.ini
@@ -16,7 +16,7 @@
;OCR engine to use. Available: "mangaocr", "mangaocrs", "glens", "bing","gvision",
;"avision", "alivetext", "azure", "winrtocr", "oneocr", "easyocr", "rapidocr",
-;"ocrspace"
+;"ocrspace", "meikiocr"
;engine =
;Local OCR engine to use for two-pass screen capture processing.
@@ -33,8 +33,8 @@
;delete_images = False
;Restrict engines to load. Available: avision,alivetext,bing,glens,gvision,azure,
-;mangaocr,mangaocrs,winrtocr,oneocr, easyocr,rapidocr,ocrspace
-;engines = avision,alivetext,bing,glens,gvision,azure,mangaocr,mangaocrs,winrtocr,oneocr,easyocr,rapidocr,ocrspace
+;mangaocr,mangaocrs,winrtocr,oneocr, easyocr,rapidocr,ocrspace,meikiocr
+;engines = avision,alivetext,bing,glens,gvision,azure,mangaocr,mangaocrs,winrtocr,oneocr,easyocr,rapidocr,ocrspace,meikiocr
;logger_format = {time:HH:mm:ss.SSS} | {message}
diff --git a/pyproject.toml b/pyproject.toml
index d6a1f0c..b672c6f 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -45,6 +45,10 @@ rapidocr = [
"onnxruntime"
]
+meikiocr = [
+ "meikiocr"
+]
+
mangaocr = [
"manga-ocr",
"setuptools<80",