easyocr

Dockerfile

@@ -1,13 +1,10 @@
 FROM python:3.12.1
 
 ENV HF_HOME /tmp
-ENV TESSDATA_PREFIX /usr/local/share/tessdata
 
 WORKDIR /code
 
-RUN apt-get update && apt-get install -y libgl1-mesa-glx libboost-thread1.74.0 tesseract-ocr tesseract-ocr-eng wget && rm -rf /var/lib/apt/lists/* && mkdir -p "${TESSDATA_PREFIX}" \
-    && echo "Created Tesseract data directory at ${TESSDATA_PREFIX}" \
-    && wget --quiet -O "${TESSDATA_PREFIX}/eng.traineddata" "https://github.com/tesseract-ocr/tessdata_best/raw/main/eng.traineddata"
+RUN apt-get update && apt-get install -y libgl1-mesa-glx libboost-thread1.74.0 && rm -rf /var/lib/apt/lists/*
 
 
 

main.py

@@ -6,8 +6,8 @@ from datetime import datetime
 from typing import List
 
 import cv2
+import easyocr
 import numpy as np
-import pytesseract
 from fastapi import FastAPI, HTTPException, Request, UploadFile
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse
@@ -51,9 +51,22 @@ app.add_middleware(
 
 # AIモデルとテンプレートを起動時に読み込む
 # trocr_pipeline = None  # load_model()のロジックをここに
+reader = None
 suit_templates = None
 
 
+@app.on_event("startup")
+def load_ocr_model():
+    """
+    アプリケーション起動時に一度だけEasyOCRのモデルを読み込み、
+    グローバル変数readerに格納する。
+    """
+    global reader
+    # 使用する言語をリストで指定（英語の場合は'en'）
+    reader = easyocr.Reader(["en"])
+    print("EasyOCR model loaded successfully.")
+
+
 # @app.on_event("startup")
 # def load_dependencies():
 #     global trocr_pipeline, suit_templates
@@ -81,14 +94,15 @@ async def analyze_image(image_paths: list[UploadFile]):
     # image_paths = request["image_paths"]
     print(image_paths)
     progress = lambda x, desc: print(x, desc)
+    global reader
     # global trocr_pipeline
     # # モデルが読み込まれているか確認
-    # if trocr_pipeline is None:
-    #     print(
-    #         "AIモデルがまだ読み込まれていません。しばらく待ってから再度お試しください。"
-    #     )
-    #     # 空の更新を返すことで、UIの状態を変えずに処理を終了
-    #     return
+    if reader is None:
+        print(
+            "AIモデルがまだ読み込まれていません。しばらく待ってから再度お試しください。"
+        )
+        # 空の更新を返すことで、UIの状態を変えずに処理を終了
+        return
 
     all_results = []
     num_total_files = len(image_paths)
@@ -187,14 +201,14 @@ async def analyze_image(image_paths: list[UploadFile]):
         try:
             ocr_results = []
             for candidate in all_candidates_global:
-                img = Image.fromarray(
-                    cv2.cvtColor(candidate["img"], cv2.COLOR_BGR2RGB)
-                )
+                # img = Image.fromarray(
+                #     cv2.cvtColor(candidate["img"], cv2.COLOR_BGR2RGB)
+                # )
                 custom_config = (
                     r"--psm 10 -c tessedit_char_whitelist=0123456789AKQJ"
                 )
 
-                text = pytesseract.image_to_string(img, config=custom_config)
+                result = reader.readtext(candidate["img"])
                 ocr_results.append(text)
             # candidates_pil_images = [
             #     Image.fromarray(cv2.cvtColor(c["img"], cv2.COLOR_BGR2RGB))
@@ -214,7 +228,7 @@ async def analyze_image(image_paths: list[UploadFile]):
         # reds = []
         for i, result in enumerate(ocr_results):
             # text = result[0]["generated_text"].upper().strip()
-            text = result
+            _, text, _ = result[0]
             print(text, is_text_valid(text))
 
             text = is_text_valid(text)