Add files

app.py

@@ -1,9 +1,12 @@
 import os
 import io
 import easyocr
+import numpy as np
 from fastapi import FastAPI, UploadFile, File
 from fastapi.responses import JSONResponse
 from PIL import Image
+from pdf2image import convert_from_bytes
+from concurrent.futures import ThreadPoolExecutor
 
 # =========================
 # EasyOCR config
@@ -16,10 +19,10 @@ os.makedirs(USER_NET_DIR, exist_ok=True)
 
 # ✅ preload reader with cached models
 reader = easyocr.Reader(
-    ['en', 'hi'],  # langs (can reduce to ['en'] if you want smaller image)
+    ['en', 'hi'],  # langs (reduce if only English needed)
     model_storage_directory=MODEL_DIR,
     user_network_directory=USER_NET_DIR,
-    download_enabled=False   # 🚫 no runtime downloads
+    download_enabled=False   # 🚫 block downloads at runtime
 )
 
 # =========================
@@ -31,26 +34,48 @@ app = FastAPI()
 async def root():
     return {"message": "OCR API is running on Hugging Face 🚀"}
 
+
+def run_ocr_on_image(image: Image.Image):
+    """Convert PIL → numpy and run OCR"""
+    image_np = np.array(image)
+    results = reader.readtext(image_np)
+    text_results = []
+    for bbox, text, prob in results:
+        text_results.append({
+            "bbox": bbox,
+            "text": text,
+            "confidence": float(prob)
+        })
+    return text_results
+
+
 @app.post("/ocr")
 async def ocr(file: UploadFile = File(...)):
     try:
-        # read image into memory
         contents = await file.read()
-        image = Image.open(io.BytesIO(contents))
 
-        # run OCR
-        results = reader.readtext(image)
+        # Detect file type
+        if file.filename.lower().endswith(".pdf"):
+            # ✅ Convert PDF to images
+            pages = convert_from_bytes(contents)
+
+            # ✅ Run OCR in parallel
+            text_results = []
+            with ThreadPoolExecutor() as executor:
+                results_list = list(executor.map(run_ocr_on_image, pages))
+                for i, page_results in enumerate(results_list, start=1):
+                    text_results.append({
+                        "page": i,
+                        "results": page_results
+                    })
 
-        # format results
-        text_results = []
-        for bbox, text, prob in results:
-            text_results.append({
-                "bbox": bbox,
-                "text": text,
-                "confidence": float(prob)
-            })
+            return JSONResponse(content={"pdf_results": text_results})
 
-        return JSONResponse(content={"results": text_results})
+        else:
+            # ✅ Normal image case
+            image = Image.open(io.BytesIO(contents))
+            text_results = run_ocr_on_image(image)
+            return JSONResponse(content={"results": text_results})
 
     except Exception as e:
         return JSONResponse(content={"error": str(e)}, status_code=500)

requirements.txt

@@ -5,5 +5,3 @@ pdf2image
 numpy
 Pillow
 python-multipart
-
-