Sync from GitHub: cff9d2103d559b06cd5ccba9969757ff860436aa

.gitignore

@@ -31,21 +31,10 @@ htmlcov/
 frontend/node_modules/
 frontend/.env.local
 
-# Documentation (keep essential ones)
 *.md
 !README_git.md
 !README.md
-DEPLOYMENT.md
-HF_DEPLOYMENT_READY.md
-IMAGE_ENHANCEMENT.md
-
-# Test files and docs
-test_*.py
-Docs/
-
-# Executables and examples
+test*
 executable.py
 client_example.py
-
-# Real-ESRGAN downloaded binaries (will be installed via Docker)
-utils/realesrgan/
+Docs

Dockerfile

@@ -2,7 +2,7 @@ FROM python:3.10-slim
 
 WORKDIR /app
 
-# Install system dependencies including Node.js and tools for Real-ESRGAN
+# Install system dependencies including Node.js
 RUN apt-get update && apt-get install -y \
     git \
     libgl1 \
@@ -12,10 +12,6 @@ RUN apt-get update && apt-get install -y \
     libxrender-dev \
     libgomp1 \
     curl \
-    wget \
-    unzip \
-    libvulkan1 \
-    libvulkan-dev \
     && curl -fsSL https://deb.nodesource.com/setup_18.x | bash - \
     && apt-get install -y nodejs \
     && rm -rf /var/lib/apt/lists/*
@@ -41,19 +37,6 @@ COPY inference.py .
 COPY app.py .
 COPY utils/ utils/
 
-# Download and setup Real-ESRGAN-ncnn-vulkan for image enhancement (after copying utils/)
-RUN mkdir -p /app/utils/realesrgan && \
-    cd /app/utils/realesrgan && \
-    wget -q https://github.com/xinntao/Real-ESRGAN-ncnn-vulkan/releases/download/v0.2.0/realesrgan-ncnn-vulkan-v0.2.0-ubuntu.zip && \
-    unzip -q realesrgan-ncnn-vulkan-v0.2.0-ubuntu.zip && \
-    rm realesrgan-ncnn-vulkan-v0.2.0-ubuntu.zip && \
-    find . -name "realesrgan-ncnn-vulkan" -type f -exec chmod +x {} \; && \
-    mv realesrgan-ncnn-vulkan-v0.2.0-ubuntu/* . 2>/dev/null || true && \
-    rmdir realesrgan-ncnn-vulkan-v0.2.0-ubuntu 2>/dev/null || true && \
-    echo "Real-ESRGAN installed at /app/utils/realesrgan/" && \
-    ls -la /app/utils/realesrgan/ && \
-    test -f /app/utils/realesrgan/realesrgan-ncnn-vulkan && echo "✓ Executable found" || echo "✗ Executable NOT found"
-
 # Expose Hugging Face Spaces default port
 EXPOSE 7860
 

app.py

@@ -98,8 +98,7 @@ async def health_check():
 @app.post("/extract")
 async def extract_invoice(
     file: UploadFile = File(..., description="Invoice image file (JPG, PNG, JPEG)"),
-    doc_id: Optional[str] = Form(None, description="Optional document identifier"),
-    enhance: Optional[bool] = Form(None, description="Enable image enhancement (default: True)")
+    doc_id: Optional[str] = Form(None, description="Optional document identifier")
 ):
     """
     Extract information from invoice image
@@ -107,7 +106,6 @@ async def extract_invoice(
     **Parameters:**
     - **file**: Invoice image file (required)
     - **doc_id**: Optional document identifier (auto-generated from filename if not provided)
-    - **enhance**: Enable image enhancement for blurry images (default: True)
     
     **Returns:**
     - JSON with extracted fields, confidence scores, and metadata
@@ -172,8 +170,8 @@ async def extract_invoice(
         if doc_id is None:
             doc_id = os.path.splitext(file.filename)[0]
         
-        # Process invoice (with optional enhancement)
-        result = InferenceProcessor.process_invoice(temp_file, doc_id, enhance=enhance)
+        # Process invoice
+        result = InferenceProcessor.process_invoice(temp_file, doc_id)
         
         # Add total request time (includes file I/O)
         result['total_request_time_sec'] = round(time.time() - request_start, 2)
@@ -201,8 +199,7 @@ async def extract_invoice(
 
 @app.post("/process-invoice")
 async def process_invoice(
-    file: UploadFile = File(..., description="Invoice image file"),
-    enhance: Optional[bool] = Form(None, description="Enable image enhancement (default: True)")
+    file: UploadFile = File(..., description="Invoice image file")
 ):
     """
     Process a single invoice and return extracted information
@@ -210,7 +207,6 @@ async def process_invoice(
     
     **Parameters:**
     - **file**: Invoice image file (required)
-    - **enhance**: Enable image enhancement for blurry images (default: True)
     
     **Returns:**
     - JSON with extracted_text, signature_coords, stamp_coords
@@ -241,8 +237,8 @@ async def process_invoice(
         # Use filename as doc_id
         doc_id = os.path.splitext(file.filename)[0] if file.filename else "invoice"
         
-        # Process invoice (with optional enhancement)
-        result = InferenceProcessor.process_invoice(temp_file, doc_id, enhance=enhance)
+        # Process invoice
+        result = InferenceProcessor.process_invoice(temp_file, doc_id)
         
         # Extract fields from result
         fields = result.get("fields", {})
@@ -307,15 +303,13 @@ async def process_invoice(
 
 @app.post("/extract_batch")
 async def extract_batch(
-    files: list[UploadFile] = File(..., description="Multiple invoice images"),
-    enhance: Optional[bool] = Form(None, description="Enable image enhancement (default: True)")
+    files: list[UploadFile] = File(..., description="Multiple invoice images")
 ):
     """
     Extract information from multiple invoice images
     
     **Parameters:**
     - **files**: List of invoice image files
-    - **enhance**: Enable image enhancement for blurry images (default: True)
     
     **Returns:**
     - JSON array with results for each invoice
@@ -350,7 +344,7 @@ async def extract_batch(
             # Process
             try:
                 doc_id = os.path.splitext(file.filename)[0]
-                result = InferenceProcessor.process_invoice(temp_file, doc_id, enhance=enhance)
+                result = InferenceProcessor.process_invoice(temp_file, doc_id)
                 results.append(result)
             except Exception as e:
                 results.append({

config.py

@@ -26,11 +26,6 @@ QUANTIZATION_CONFIG = {
 # Image processing settings
 MAX_IMAGE_SIZE = 800  # Maximum dimension for resizing
 
-# Image Enhancement Settings (Real-ESRGAN)
-ENABLE_IMAGE_ENHANCEMENT = True  # Enable/disable image enhancement
-ENHANCEMENT_SCALE = 2  # Upscaling factor (2, 3, or 4)
-ENHANCEMENT_MODEL = "realesrgan-x4plus"  # Model: realesrgan-x4plus, realesrgan-x4plus-anime, realesrnet-x4plus
-
 # Detection thresholds
 YOLO_CONFIDENCE_THRESHOLD = 0.25
 

frontend/src/components/ResultCard.jsx

@@ -1,5 +1,5 @@
 import React, { useRef, useEffect, useState } from 'react';
-import { SlidersHorizontal, Download, Eye } from 'lucide-react';
+import { SlidersHorizontal } from 'lucide-react';
 
 const ResultCard = ({ result, imageData, processedImageData, onReprocess, isProcessing }) => {
   const canvasRef = useRef(null);
@@ -11,7 +11,6 @@ const ResultCard = ({ result, imageData, processedImageData, onReprocess, isProc
   const [adjustedDataUrl, setAdjustedDataUrl] = useState(null);
   const [previewDimensions, setPreviewDimensions] = useState({ width: 0, height: 0 });
   const [currentImageData, setCurrentImageData] = useState(processedImageData || imageData);
-  const [showEnhanced, setShowEnhanced] = useState(false);
 
   // Function to crop image regions
   const cropRegion = (img, coords, scaleX, scaleY) => {
@@ -254,43 +253,6 @@ const ResultCard = ({ result, imageData, processedImageData, onReprocess, isProc
               <span>100% (Best)</span>
             </div>
           </div>
-
-          {/* Enhanced Image Toggle & Download */}
-          {result.image_enhanced && result.enhanced_image && (
-            <div className="bg-green-50 rounded-lg p-4 shadow-sm border border-green-200">
-              <div className="flex items-center justify-between mb-2">
-                <div className="flex items-center gap-2">
-                  <svg className="w-5 h-5 text-green-600" fill="none" viewBox="0 0 24 24" stroke="currentColor">
-                    <path strokeLinecap="round" strokeLinejoin="round" strokeWidth={2} d="M9 12l2 2 4-4m6 2a9 9 0 11-18 0 9 9 0 0118 0z" />
-                  </svg>
-                  <span className="text-sm font-medium text-green-800">Image Enhanced (2x)</span>
-                </div>
-                <div className="flex gap-2">
-                  <button
-                    onClick={() => {
-                      const enhancedData = `data:image/png;base64,${result.enhanced_image}`;
-                      setCurrentImageData(showEnhanced ? imageData : enhancedData);
-                      setShowEnhanced(!showEnhanced);
-                    }}
-                    className="flex items-center gap-1 px-3 py-1 bg-green-600 hover:bg-green-700 text-white rounded text-xs font-medium transition-colors"
-                  >
-                    <Eye className="w-3 h-3" />
-                    {showEnhanced ? 'Show Original' : 'Show Enhanced'}
-                  </button>
-                  <a
-                    href={`data:image/png;base64,${result.enhanced_image}`}
-                    download={`enhanced_${result.filename || 'invoice'}.png`}
-                    className="flex items-center gap-1 px-3 py-1 bg-blue-600 hover:bg-blue-700 text-white rounded text-xs font-medium transition-colors"
-                  >
-                    <Download className="w-3 h-3" />
-                    Download
-                  </a>
-                </div>
-              </div>
-              <p className="text-xs text-green-700">Real-ESRGAN enhancement applied before processing</p>
-            </div>
-          )}
-          
           <div className="relative bg-gray-50 rounded-lg p-4 flex justify-center items-center">
             <canvas ref={canvasRef} className="max-w-full h-auto rounded shadow-md" />
             {isProcessing && (

inference.py

@@ -7,7 +7,6 @@ import time
 import json
 import codecs
 import re
-import os
 from PIL import Image
 from qwen_vl_utils import process_vision_info
 from typing import Dict, Tuple
@@ -16,13 +15,9 @@ from config import (
     MAX_IMAGE_SIZE,
     HP_VALID_RANGE,
     ASSET_COST_VALID_RANGE,
-    COST_PER_GPU_HOUR,
-    ENABLE_IMAGE_ENHANCEMENT,
-    ENHANCEMENT_SCALE,
-    ENHANCEMENT_MODEL
+    COST_PER_GPU_HOUR
 )
 from model_manager import model_manager
-from utils.image_enhancer import get_enhancer
 
 
 EXTRACTION_PROMPT = """
@@ -69,47 +64,18 @@ class InferenceProcessor:
     """Handles VLM inference, validation, and result processing"""
     
     @staticmethod
-    def preprocess_image(image_path: str, enhance: bool = None) -> Tuple[Image.Image, str]:
-        """Load, enhance (optional), and resize image if needed
+    def preprocess_image(image_path: str) -> Image.Image:
+        """Load and resize image if needed"""
+        image = Image.open(image_path).convert("RGB")
         
-        Args:
-            image_path: Path to input image
-            enhance: Whether to enhance image quality before processing (None=use config default)
-        
-        Returns:
-            Tuple of (PIL Image ready for VLM, path to image file for YOLO)
-        """
-        # Use config default if not specified
-        if enhance is None:
-            enhance = ENABLE_IMAGE_ENHANCEMENT
-        
-        # Step 1: Enhance image if enabled
-        enhanced_path = image_path
-        
-        if enhance:
-            try:
-                enhancer = get_enhancer()
-                enhanced_path = enhancer.enhance_image(
-                    image_path, 
-                    scale=ENHANCEMENT_SCALE,
-                    model_name=ENHANCEMENT_MODEL
-                )
-            except Exception as e:
-                print(f"⚠️ Enhancement failed: {str(e)}, using original image")
-                enhanced_path = image_path
-        
-        # Step 2: Load image
-        image = Image.open(enhanced_path).convert("RGB")
-        
-        # Step 3: Resize if too large
+        # Resize if too large
         if max(image.size) > MAX_IMAGE_SIZE:
             ratio = MAX_IMAGE_SIZE / max(image.size)
             new_size = (int(image.size[0] * ratio), int(image.size[1] * ratio))
             image = image.resize(new_size, Image.LANCZOS)
             print(f"🔄 Image resized to {new_size}")
         
-        # Return both PIL Image and path (path will be cleaned up by caller)
-        return image, enhanced_path
+        return image
     
     @staticmethod
     def run_vlm_extraction(image: Image.Image) -> Tuple[str, float]:
@@ -318,14 +284,13 @@ class InferenceProcessor:
         return validated, field_confidence, warnings
     
     @staticmethod
-    def process_invoice(image_path: str, doc_id: str = None, enhance: bool = None) -> Dict:
+    def process_invoice(image_path: str, doc_id: str = None) -> Dict:
         """
         Complete invoice processing pipeline
         
         Args:
             image_path: Path to invoice image
             doc_id: Document identifier (optional)
-            enhance: Whether to enhance image (None=use config default)
         
         Returns:
             dict: Complete JSON output with all fields
@@ -338,27 +303,14 @@ class InferenceProcessor:
             import os
             doc_id = os.path.splitext(os.path.basename(image_path))[0]
         
-        # Step 1: Preprocess image (with optional enhancement)
+        # Step 1: Preprocess image
         t1 = time.time()
-        image, enhanced_image_path = InferenceProcessor.preprocess_image(image_path, enhance=enhance)
+        image = InferenceProcessor.preprocess_image(image_path)
         timing_breakdown['image_preprocessing'] = round(time.time() - t1, 3)
         
-        # Save enhanced image info for response
-        image_was_enhanced = (enhanced_image_path != image_path)
-        enhanced_image_base64 = None
-        
-        if image_was_enhanced:
-            # Convert enhanced image to base64 for response
-            import base64
-            try:
-                with open(enhanced_image_path, 'rb') as f:
-                    enhanced_image_base64 = base64.b64encode(f.read()).decode('utf-8')
-            except:
-                pass
-        
-        # Step 2: YOLO Detection (use enhanced image path for consistency)
+        # Step 2: YOLO Detection
         t2 = time.time()
-        signature_info, stamp_info, signature_conf, stamp_conf = model_manager.detect_sign_stamp(enhanced_image_path)
+        signature_info, stamp_info, signature_conf, stamp_conf = model_manager.detect_sign_stamp(image_path)
         timing_breakdown['yolo_detection'] = round(time.time() - t2, 3)
         
         # Step 3: VLM Extraction
@@ -366,17 +318,10 @@ class InferenceProcessor:
         vlm_output, vlm_latency = InferenceProcessor.run_vlm_extraction(image)
         timing_breakdown['vlm_inference'] = round(vlm_latency, 3)
         
-        # Clean up image and enhanced file if it was created
+        # Clean up image
         image.close()
         del image
         
-        # Cleanup enhanced temp file if created
-        if enhanced_image_path != image_path:
-            try:
-                os.unlink(enhanced_image_path)
-            except:
-                pass
-        
         # Step 4: Parse JSON
         t4 = time.time()
         raw_json = InferenceProcessor.extract_json_from_output(vlm_output)
@@ -412,9 +357,7 @@ class InferenceProcessor:
             "processing_time_sec": round(total_time, 2),
             "timing_breakdown": timing_breakdown,
             "cost_estimate_usd": round(cost_estimate, 6),
-            "warnings": warnings if warnings else None,
-            "image_enhanced": image_was_enhanced,
-            "enhanced_image": enhanced_image_base64 if image_was_enhanced else None
+            "warnings": warnings if warnings else None
         }
         
         return result