From 79891063ddb721a350c4f6d26be9e0131058a891 Mon Sep 17 00:00:00 2001
From: Benjamin Admin <benjaminadmin@MacBookPro.fritz.box>
Date: Thu, 12 Mar 2026 13:32:54 +0100
Subject: [PATCH] fix: pin PaddlePaddle 2.6.2 + PaddleOCR 2.8.1 (stable, no PIR
 bug)

PaddlePaddle 3.x hat oneDNN/PIR Executor Bug. Zurueck auf 2.6.2
mit bewaeherter ocr() API statt predict().

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 paddleocr-service/main.py          | 53 +++++++++++++-----------------
 paddleocr-service/requirements.txt |  4 +--
 2 files changed, 25 insertions(+), 32 deletions(-)

diff --git a/paddleocr-service/main.py b/paddleocr-service/main.py
index 086d762..f66aef7 100644
--- a/paddleocr-service/main.py
+++ b/paddleocr-service/main.py
@@ -1,4 +1,4 @@
-"""PaddleOCR Remote Service — PP-OCRv5 Latin auf x86_64."""
+"""PaddleOCR Remote Service — PP-OCRv4 Latin auf x86_64."""
 
 import io
 import logging
@@ -23,12 +23,11 @@ def get_engine():
     if _engine is None:
         from paddleocr import PaddleOCR
 
-        logger.info("Loading PaddleOCR model (first time may download)...")
+        logger.info("Loading PaddleOCR model...")
         _engine = PaddleOCR(
-            text_recognition_model_name="latin_PP-OCRv5_mobile_rec",
-            use_doc_orientation_classify=False,
-            use_doc_unwarping=False,
-            use_textline_orientation=False,
+            lang="latin",
+            use_angle_cls=True,
+            show_log=False,
         )
         logger.info("PaddleOCR model loaded successfully")
     return _engine
@@ -49,7 +48,7 @@ def startup_load_model():
 @app.get("/health")
 def health():
     if _ready:
-        return {"status": "ok", "model": "PP-OCRv5-latin"}
+        return {"status": "ok", "model": "PP-OCRv4-latin"}
     return {"status": "loading"}
 
 
@@ -69,31 +68,25 @@ async def ocr(
     img_np = np.array(img)
 
     engine = get_engine()
-    result = engine.predict(img_np)
+    result = engine.ocr(img_np)
 
     words = []
-    for item in result:
-        rec_texts = item.get("rec_texts", [])
-        rec_scores = item.get("rec_scores", [])
-        dt_polys = item.get("dt_polys", [])
-
-        for text, score, poly in zip(rec_texts, rec_scores, dt_polys):
-            if not text or not text.strip():
-                continue
-            xs = [p[0] for p in poly]
-            ys = [p[1] for p in poly]
-            x_min, x_max = min(xs), max(xs)
-            y_min, y_max = min(ys), max(ys)
-            words.append(
-                {
-                    "text": text.strip(),
-                    "left": int(x_min),
-                    "top": int(y_min),
-                    "width": int(x_max - x_min),
-                    "height": int(y_max - y_min),
-                    "conf": round(float(score) * 100, 1),
-                }
-            )
+    for line in result[0] or []:
+        box, (text, conf) = line[0], line[1]
+        x_min = min(p[0] for p in box)
+        y_min = min(p[1] for p in box)
+        x_max = max(p[0] for p in box)
+        y_max = max(p[1] for p in box)
+        words.append(
+            {
+                "text": text.strip(),
+                "left": int(x_min),
+                "top": int(y_min),
+                "width": int(x_max - x_min),
+                "height": int(y_max - y_min),
+                "conf": round(conf * 100, 1),
+            }
+        )
 
     return {
         "words": words,
diff --git a/paddleocr-service/requirements.txt b/paddleocr-service/requirements.txt
index 47f8951..0d3e063 100644
--- a/paddleocr-service/requirements.txt
+++ b/paddleocr-service/requirements.txt
@@ -1,5 +1,5 @@
-paddlepaddle>=3.0.0
-paddleocr>=2.9.0
+paddlepaddle==2.6.2
+paddleocr==2.8.1
 fastapi>=0.110.0
 uvicorn>=0.25.0
 python-multipart>=0.0.6