kha-white · Icekey · Oct 21, 2023 · Jun 16, 2024 · Feb 4, 2025
diff --git a/README.md b/README.md
@@ -34,6 +34,11 @@ Run in command line:
 pip3 install manga-ocr
 ```
 
+Local install
+```commandline
+pip install -e C:\Users\bluek\Documents\Python\manga-ocr
+```
+
 ## Troubleshooting
 
 - `ImportError: DLL load failed while importing fugashi: The specified module could not be found.` - might be because of Python installed from Microsoft Store, try installing Python from the [official site](https://www.python.org/downloads)

diff --git a/manga_ocr/__init__.py b/manga_ocr/__init__.py
@@ -1,3 +1,3 @@
-__version__ = '0.1.11'
+__version__ = '0.1.12'
 
 from manga_ocr.ocr import MangaOcr
diff --git a/manga_ocr/__main__.py b/manga_ocr/__main__.py
@@ -4,7 +4,7 @@
 
 
 def main():
-    fire.Fire(run)
+    fire.Fire(run(read_from='cli', write_to='cli'))
 
 
 if __name__ == '__main__':

diff --git a/manga_ocr/ocr.py b/manga_ocr/ocr.py
@@ -17,12 +17,15 @@ def __init__(self, pretrained_model_name_or_path='kha-white/manga-ocr-base', for
 
         if not force_cpu and torch.cuda.is_available():
             logger.info('Using CUDA')
+            print('Using CUDA')
             self.model.cuda()
         elif not force_cpu and torch.backends.mps.is_available():
             logger.info('Using MPS')
+            print('Using MPS')
             self.model.to('mps')
         else:
             logger.info('Using CPU')
+            print('Using CPU')
 
         example_path = Path(__file__).parent / 'assets/example.jpg'
         if not example_path.is_file():
@@ -33,18 +36,24 @@ def __init__(self, pretrained_model_name_or_path='kha-white/manga-ocr-base', for
 
     def __call__(self, img_or_path):
         if isinstance(img_or_path, str) or isinstance(img_or_path, Path):
-            img = Image.open(img_or_path)
+            img = [Image.open(img_or_path)]
         elif isinstance(img_or_path, Image.Image):
+            img = [img_or_path]
+        elif type(img_or_path) in (tuple, list):
             img = img_or_path
         else:
             raise ValueError(f'img_or_path must be a path or PIL.Image, instead got: {img_or_path}')
 
-        img = img.convert('L').convert('RGB')
+        img = [i.convert('L').convert('RGB') for i in img]
 
         x = self._preprocess(img)
-        x = self.model.generate(x[None].to(self.model.device), max_length=300)[0].cpu()
-        x = self.tokenizer.decode(x, skip_special_tokens=True)
-        x = post_process(x)
+
+        if x.dim() == 3:
+            x = x.unsqueeze(0)
+
+        x = self.model.generate(x.to(self.model.device), max_length=300).cpu()
+        x = [self.tokenizer.decode(x[i], skip_special_tokens=True) for i in range(x.size(dim=0))]
+        x = [post_process(i) for i in x]
         return x
 
     def _preprocess(self, img):
@@ -58,4 +67,6 @@ def post_process(text):
     text = re.sub('[・.]{2,}', lambda x: (x.end() - x.start()) * '.', text)
     text = jaconv.h2z(text, ascii=True, digit=True)
 
+    if not bool(text.strip()):
+        return "<no ocr>"
     return text
diff --git a/manga_ocr/run.py b/manga_ocr/run.py
@@ -31,10 +31,13 @@ def process_and_write_results(mocr, img_or_path, write_to):
 
     if write_to == 'clipboard':
         pyperclip.copy(text)
+    elif write_to == 'cli':
+        for i in text:
+            print(i)
     else:
         write_to = Path(write_to)
         if write_to.suffix != '.txt':
-            raise ValueError('write_to must be either "clipboard" or a path to a text file')
+            raise ValueError('write_to must be either "clipboard", "cli" or a path to a text file')
 
         with write_to.open('a', encoding="utf-8") as f:
             f.write(text + '\n')
@@ -101,11 +104,19 @@ def run(read_from='clipboard',
                     process_and_write_results(mocr, img, write_to)
 
             time.sleep(delay_secs)
+    elif read_from == 'cli':
+        logger.info(f'Reading from cli')
 
+        while True:
+            path = input('Enter image path:\n')
+            paths = path.split(",")
+            images = get_images(paths)
+            if len(images) > 0:
+                process_and_write_results(mocr, images, write_to)
     else:
         read_from = Path(read_from)
         if not read_from.is_dir():
-            raise ValueError('read_from must be either "clipboard" or a path to a directory')
+            raise ValueError('read_from must be either "clipboard", "cli" or a path to a directory')
 
         logger.info(f'Reading from directory {read_from}')
 
@@ -130,5 +141,21 @@ def run(read_from='clipboard',
             time.sleep(delay_secs)
 
 
+def get_images(paths):
+    images = []
+    for path in paths:
+        read_from = Path(path)
+        if not read_from.is_file():
+            print(f'{path} is not a file')
+            continue
+        try:
+            img = Image.open(read_from)
+            img.load()
+            images.append(img)
+        except (UnidentifiedImageError, OSError) as e:
+            print(f'Error while reading file {read_from}: {e}')
+    return images
+
+
 if __name__ == '__main__':
     fire.Fire(run)
diff --git a/setup.py b/setup.py
@@ -5,7 +5,7 @@
 
 setup(
     name="manga-ocr",
-    version='0.1.11',
+    version='0.1.12',
     description="OCR for Japanese manga",
     long_description=long_description,
     long_description_content_type="text/markdown",