Skip to content
The Chat Archive
claude-3-7-sonnet-20250219· Anthropic LMArena Human Preference 140K

[code] [code]

  1. User
    # clipcrop.py
    import asyncio
    import os
    from logging import getLogger
    from typing import Optional
    
    import cv2
    import numpy as np
    import pytesseract
    import torch
    from PIL import Image
    from ultralytics import YOLO
    
    from configs.path_config import EXAMPLE_IMAGE, YOLO_MODELS_DIR
    
    logger_local = getLogger(__name__)
    
    
    class ClipCropAsync:
        def __init__(self, image_path: str, yolo_model_path: str = YOLO_MODELS_DIR / "yolov5su.pt"):
            self.image_path = str(image_path)
            self.yolo_model_path = yolo_model_path
            self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
            self.model: Optional[YOLO] = None
    
        async def _load_model(self) -> None:
            # モデルファイルが存在しない場合は、ultralytics が自動でダウンロード
            if not os.path.exists(self.yolo_model_path):
                logger_local.info(f"YOLO model not found at {self.yolo_model_path}, downloading...")
            self.model = YOLO(self.yolo_model_path)
            # 必要なら GPU に転送
            try:
                self.model.to(self.device)
            except Exception:
                pass
    
        async def auto_captcha(self, threshold: int = 3, tx: str = "default") -> Image.Image:
            if self.model is None:
                await self._load_model()
    
            # 画像読み込み
            image = await asyncio.to_thread(cv2.imread, self.image_path)
            if image is None:
                raise FileNotFoundError(f"Image not found or cannot be read: {self.image_path}")
    
            # グレースケール変換
            gray = await asyncio.to_thread(cv2.cvtColor, image, cv2.COLOR_BGR2GRAY)
    
            # OCR 前処理
            _, thresh_bin = cv2.threshold(gray, 200, 255, cv2.THRESH_BINARY)
            kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (1, 1))
            close = cv2.erode(thresh_bin, kernel, iterations=2)
            invert = cv2.bitwise_not(close)
    
            inf = await asyncio.to_thread(pytesseract.image_to_string, invert)
            lines = [x for x in inf.split('\n') if len(x) > 2]
            txt = lines[1] if len(lines) > 1 else tx
            logger_local.info(f"Extracted text: {txt}")
    
            # 画像シャープ化
            blur = cv2.medianBlur(gray, 3)
            sharpen_kernel = np.array([[-1, -1, -1],
                                       [-1, 9, -1],
                                       [-1, -1, -1]])
            sharpen = cv2.filter2D(blur, -1, sharpen_kernel)
    
            # 二値化・モルフォロジー
            _, thresh_inv = cv2.threshold(sharpen, 160, 255, cv2.THRESH_BINARY_INV)
            close2 = cv2.morphologyEx(thresh_inv, cv2.MORPH_CLOSE, kernel, iterations=2)
    
            # 輪郭抽出・描画
            cnts, _ = cv2.findContours(close2, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
            for c in cnts:
                x, y, w, h = cv2.boundingRect(c)
                if w == h and w > 5:
                    cv2.rectangle(image, (x, y), (x + w, y + h), (36, 255, 12), 2)
    
            # BGR→RGB 変換して PIL Image 化
            rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
            return Image.fromarray(rgb)
    
        async def extract_image(self, text: str, num: int = 3) -> list[dict]:
            if self.model is None:
                await self._load_model()
    
            # 画像読み込み
            image = await asyncio.to_thread(cv2.imread, self.image_path)
            if image is None:
                raise FileNotFoundError(f"Image not found or cannot be read: {self.image_path}")
    
            # モデル推論(スレッドプールで実行)
            results = await asyncio.to_thread(self.model, image)
            boxes = results[0].boxes.xyxy.cpu().numpy()
    
            # 面積でソートして上位 num 件を抽出
            sorted_boxes = sorted(
                boxes,
                key=lambda b: (b[2] - b[0]) * (b[3] - b[1]),
                reverse=True
            )[:num]
    
            crops: list[dict] = []
            for xmin, ymin, xmax, ymax in sorted_boxes:
                crop_img = image[int(ymin):int(ymax), int(xmin):int(xmax)]
                # BGR→RGB
                pil_img = Image.fromarray(crop_img[:, :, ::-1])
                crops.append({
                    "image": pil_img,
                    "bbox": (int(xmin), int(ymin), int(xmax), int(ymax))
                })
            return crops
    
        # crop_to_prompt は extract_image のエイリアスとして残します
        async def crop_to_prompt(self, prompt: str, num: int = 1) -> list[dict]:
            return await self.extract_image(prompt, num=num)
    
    
    if __name__ == "__main__":
        from logging.config import dictConfig
        from configs.logging_config import LOGGING_CONFIG
    
        dictConfig(LOGGING_CONFIG)
    
        async def main():
            # EXAMPLE_IMAGE のテスト
            cc = ClipCropAsync(EXAMPLE_IMAGE)
    
            # キャプチャ解読テスト
            captcha_result = await cc.auto_captcha()
            captcha_result.save("captcha_result.png")
            logger_local.info("Saved captcha result to captcha_result.png")
    
            # プロンプトに基づくクロップテスト
            crops = await cc.crop_to_prompt("object", num=3)
            for i, crop in enumerate(crops):
                filename = f"crop_{i}.png"
                crop["image"].save(filename)
                logger_local.info(f"Crop {i}: bbox={crop['bbox']}, saved to {filename}")
    
        asyncio.run(main())
    
    C:\Users\faker\Desktop\lora_test\.venv\Scripts\python.exe C:\Users\faker\Desktop\lora_test\src\utils\clipcrop.py 
    Traceback (most recent call last):
      File "C:\Users\faker\Desktop\lora_test\.venv\Lib\site-packages\pytesseract\pytesseract.py", line 275, in run_tesseract
        proc = subprocess.Popen(cmd_args, **subprocess_args())
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 1026, in __init__
        self._execute_child(args, executable, preexec_fn, close_fds,
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\subprocess.py", line 1538, in _execute_child
        hp, ht, pid, tid = _winapi.CreateProcess(executable, args,
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
    FileNotFoundError: [WinError 2] 指定されたファイルが見つかりません。
    
    During handling of the above exception, another exception occurred:
    
    Traceback (most recent call last):
      File "C:\Users\faker\Desktop\lora_test\src\utils\clipcrop.py", line 140, in <module>
        asyncio.run(main())
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\asyncio\runners.py", line 194, in run
        return runner.run(main)
               ^^^^^^^^^^^^^^^^
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\asyncio\runners.py", line 118, in run
        return self._loop.run_until_complete(task)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\asyncio\base_events.py", line 687, in run_until_complete
        return future.result()
               ^^^^^^^^^^^^^^^
      File "C:\Users\faker\Desktop\lora_test\src\utils\clipcrop.py", line 129, in main
        captcha_result = await cc.auto_captcha()
                         ^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\Desktop\lora_test\src\utils\clipcrop.py", line 55, in auto_captcha
        inf = await asyncio.to_thread(pytesseract.image_to_string, invert)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\asyncio\threads.py", line 25, in to_thread
        return await loop.run_in_executor(None, func_call)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\AppData\Local\Programs\Python\Python312\Lib\concurrent\futures\thread.py", line 58, in run
        result = self.fn(*self.args, **self.kwargs)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\Desktop\lora_test\.venv\Lib\site-packages\pytesseract\pytesseract.py", line 486, in image_to_string
        return {
               ^
      File "C:\Users\faker\Desktop\lora_test\.venv\Lib\site-packages\pytesseract\pytesseract.py", line 489, in <lambda>
        Output.STRING: lambda: run_and_get_output(*args),
                               ^^^^^^^^^^^^^^^^^^^^^^^^^
      File "C:\Users\faker\Desktop\lora_test\.venv\Lib\site-packages\pytesseract\pytesseract.py", line 352, in run_and_get_output
        run_tesseract(**kwargs)
      File "C:\Users\faker\Desktop\lora_test\.venv\Lib\site-packages\pytesseract\pytesseract.py", line 280, in run_tesseract
        raise TesseractNotFoundError()
    pytesseract.pytesseract.TesseractNotFoundError: tesseract is not installed or it's not in your PATH. See README file for more information.
    
    Process finished with exit code 1
    
  2. Assistant

    It looks like you're encountering an error with the clipcrop.py script. The main issue is that Tesseract OCR is not installed or not properly configured in your PATH. This is indicated by the error message:

    pytesseract.pytesseract.TesseractNotFoundError: tesseract is not installed or it's not in your PATH. See README file for more information.
    

    To fix this issue:

    1. Install Tesseract OCR on your Windows system:

    2. Add Tesseract to your PATH:

      • Right-click on "This PC" or "My Computer" and select "Properties"
      • Click on "Advanced system settings"
      • Click on "Environment Variables"
      • Under "System variables", find and select "Path", then click "Edit"
      • Click "New" and add the Tesseract installation directory (e.g., C:\Program Files\Tesseract-OCR)
      • Click "OK" to close all dialogs
    3. Configure pytesseract in your code (alternative approach):

      • You can also explicitly set the Tesseract executable path in your code:
      pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
      
      • Add this line near the top of your script after importing pytesseract

    After installing Tesseract and configuring the path, your script should be able to run without the Tesseract-related error.

    Would you like me to explain or break down any part of the code in more detail?

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.