app.py

Dockerfile

@@ -0,0 +1,31 @@
+# Use Python 3.9 as the base image 
+FROM python:3.9
+
+# Create a user with ID 1000
+RUN useradd -m -u 1000 user
+USER user
+ENV PATH="/home/user/.local/bin:$PATH"
+
+# Set the working directory
+WORKDIR /app
+
+# Copy the requirements.txt file and install dependencies
+COPY --chown=user ./requirements.txt requirements.txt
+RUN pip install --no-cache-dir --upgrade -r requirements.txt
+
+# Install system dependencies needed for OpenCV, Tesseract, and other libraries
+USER root
+RUN apt-get update && apt-get install -y \
+    libgl1-mesa-glx \
+    libglib2.0-0 \
+    tesseract-ocr \
+    && rm -rf /var/lib/apt/lists/*
+
+# Return to the user 'user' for safety
+USER user
+
+# Copy the entire app directory to the container
+COPY --chown=user . /app
+
+# Set the command to run the app with Gunicorn
+CMD ["gunicorn", "-b", "0.0.0.0:7860", "app:app"]

app.py

@@ -0,0 +1,244 @@
+import os
+import csv
+import torch
+import torchvision
+import easyocr
+import shutil
+import random
+import cv2
+from glob import glob
+from ultralytics import YOLOv10
+import random
+from glob import glob
+from ultralytics import YOLOv10
+import supervision as sva
+from ultralytics import YOLOv10
+import supervision as sv
+import supervision as sv
+from flask import Flask, request, jsonify, send_from_directory, render_template
+import pytesseract
+# Set the tesseract_cmd based on the operating system
+if os.name == 'nt':  # For Windows
+    pytesseract.pytesseract.tesseract_cmd = r'C:\Program Files\Tesseract-OCR\tesseract.exe'
+elif os.name == 'posix':  # For Linux/MacOS
+    pytesseract.pytesseract.tesseract_cmd = '/usr/bin/tesseract'  # Common location for Tesseract on Linux/MacOS
+
+import textwrap
+app = Flask(__name__)
+
+def enhance_contrast(image):
+    gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
+    equalized_image = cv2.equalizeHist(gray_image)
+    return equalized_image
+
+
+def calculate_iou(bbox1, bbox2):
+    x1_max = max(bbox1[0], bbox2[0])
+    y1_max = max(bbox1[1], bbox2[1])
+    x2_min = min(bbox1[2], bbox2[2])
+    y2_min = min(bbox1[3], bbox2[3])
+
+    inter_area = max(0, x2_min - x1_max) * max(0, y2_min - y1_max)
+
+    bbox1_area = (bbox1[2] - bbox1[0]) * (bbox1[3] - bbox1[1])
+    bbox2_area = (bbox2[2] - bbox2[0]) * (bbox2[3] - bbox2[1])
+
+    iou = inter_area / float(bbox1_area + bbox2_area - inter_area) if (bbox1_area + bbox2_area - inter_area) > 0 else 0
+    return iou
+
+
+def recreate_directories():
+    # Paths for the directories
+    cropped_dir = "./app/cropped_images/"
+    output_dir1 = "./app/Folder1"
+    output_dir2 = "./app/Folder2"
+    output_dir3 = "./app/Folder3"
+    UPLOAD_FOLDER = "./app/data1"
+
+    # Remove existing directories
+    if os.path.exists(cropped_dir):
+        shutil.rmtree(cropped_dir)
+    if os.path.exists(output_dir1):
+        shutil.rmtree(output_dir1)
+    if os.path.exists(output_dir2):
+        shutil.rmtree(output_dir2)
+    if os.path.exists(output_dir3):
+        shutil.rmtree(output_dir3)
+    if os.path.exists(UPLOAD_FOLDER):
+        shutil.rmtree(UPLOAD_FOLDER)
+
+    # Recreate directories
+    os.makedirs(cropped_dir, exist_ok=True)
+    os.makedirs(output_dir1, exist_ok=True)
+    os.makedirs(output_dir2, exist_ok=True)
+    os.makedirs(output_dir3, exist_ok=True)
+    os.makedirs(UPLOAD_FOLDER, exist_ok=True)
+
+@app.route('/')
+def index():
+    return render_template('index3.html')  # This will serve your HTML page
+
+@app.route('/upload', methods=['POST'])
+def upload_file():
+    recreate_directories()
+    if 'invoice-upload' not in request.files:
+        return jsonify({'error': 'No file part'}), 400
+    file = request.files['invoice-upload']
+    if file.filename == '':
+        return jsonify({'error': 'No selected file'}), 400
+    if file:
+        file_path = os.path.join('./app/data1', file.filename)
+        file.save(file_path)
+        output_image, output_csv,image_path = process_image()
+
+        return jsonify({
+            'image_path': output_image,
+            'csv_path': output_csv,
+            'original_image': image_path
+        })
+def process_image():
+    print("Current working directory:", os.getcwd())
+    
+    # Check contents in the root directory
+    print("Current directory contents:", os.listdir('/'))
+
+    model = YOLOv10(f'./runs/detect/train3/weights/best.pt')
+    #device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    #model.to(device)
+
+    dataset = sv.DetectionDataset.from_yolo(
+        images_directory_path=f"./data/MyNewVersion5.0Dataset/valid/images",
+        annotations_directory_path=f"./data/MyNewVersion5.0Dataset/valid/labels",
+        data_yaml_path=f"./data/MyNewVersion5.0Dataset/data.yaml"
+    )
+    bounding_box_annotator = sv.BoundingBoxAnnotator()
+    label_annotator = sv.LabelAnnotator()
+    image_dir = "./app/data1"
+    files = os.listdir('./app/data1')
+    files.sort()
+    files = files[0:100]
+    print(files)
+    counter = 0
+    for ii in files:
+        random_image_data = cv2.imread('./app/data1/' + ii)
+        random_image_data1 = cv2.imread('./app/data1/' + ii)
+        results = model(source='./app/data1/' + ii, conf=0.07)[0]
+        detections = sv.Detections.from_ultralytics(results)
+        annotated_image = bounding_box_annotator.annotate(scene=random_image_data, detections=detections)
+        annotated_image = label_annotator.annotate(scene=annotated_image, detections=detections)
+        save_path = "./app/Folder1/" + "detection" + ii
+        cv2.imwrite(save_path, annotated_image)
+        print(f"Annotated image saved at {save_path}")
+        bounding_boxes = results.boxes.xyxy.cpu().numpy()
+        class_ids = results.boxes.cls.cpu().numpy()
+        confidences = results.boxes.conf.cpu().numpy()
+        bounding_box_save_path = "./bounding_boxes.txt"
+        with open(bounding_box_save_path, 'w') as f:
+            for i, (bbox, class_id, confidence) in enumerate(zip(bounding_boxes, class_ids, confidences)):
+                x1, y1, x2, y2 = map(int, bbox)
+                f.write(f"Object {i + 1}: Class {class_id}, Confidence: {confidence:.2f}, "
+                        f"Bounding box: ({x1}, {y1}, {x2}, {y2})\n")
+                cropped_image = random_image_data1[y1:y2, x1:x2]
+                cropped_image_path = os.path.join('./app/cropped_images/', f"cropped_object_{i + 1}.jpg")
+                cv2.imwrite(cropped_image_path, cropped_image)
+                print(f"Enhanced cropped image saved at {cropped_image_path}")
+        print(f"Checking contents of /app/data: {bounding_box_save_path}")
+        print(f"Directory listing: {os.listdir('./app/Folder1')}")
+        print(f"Bounding box coordinates saved at {bounding_box_save_path}")
+        import re
+        input_file_path = './bounding_boxes.txt'
+        cropped_images_folder = './app/cropped_images/'
+        output_csv_path = './app/Folder2/' + ii + 'bounding_boxes_with_recognition.csv'
+        with open(input_file_path, 'r') as infile:
+            lines = infile.readlines()
+        with open(output_csv_path, 'w', newline='', encoding='utf-8') as csvfile:
+            csv_writer = csv.writer(csvfile)
+            csv_writer.writerow(['Object ID', 'Bounding Box', 'Image Name', 'Recognized Text'])
+            for i, line in enumerate(lines):
+                object_id = f"Object_{i + 1}"
+                bounding_box_info = line.strip()
+                cropped_image_name = f"cropped_object_{i + 1}.jpg"
+                cropped_image_path = os.path.join(cropped_images_folder, cropped_image_name)
+                if os.path.exists(cropped_image_path):
+                    bbox_match = re.search(r"Bounding box: \((\d+), (\d+), (\d+), (\d+)\)", bounding_box_info)
+                    if bbox_match:
+                        x1, y1, x2, y2 = map(int, bbox_match.groups())
+                        detected_boxes = [[x1, x2, y1, y2]]
+                    else:
+                        print("No bounding box found in the info.")
+                    cropped_image = cv2.imread(cropped_image_path, cv2.IMREAD_GRAYSCALE)
+                    recognized_text = pytesseract.image_to_string(cropped_image,config="--psm 6 -c tessedit_char_whitelist=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz-")
+                    print(f"Recognized Text: {recognized_text}")
+                   
+                csv_writer.writerow([object_id, bounding_box_info, cropped_image_name, recognized_text])
+        print(f"CSV file with recognition results saved at {output_csv_path}")
+
+        def calculate_iou(bbox1, bbox2):
+            x1_max = max(bbox1[0], bbox2[0])
+            y1_max = max(bbox1[1], bbox2[1])
+            x2_min = min(bbox1[2], bbox2[2])
+            y2_min = min(bbox1[3], bbox2[3])
+
+            inter_area = max(0, x2_min - x1_max) * max(0, y2_min - y1_max)
+
+            bbox1_area = (bbox1[2] - bbox1[0]) * (bbox1[3] - bbox1[1])
+            bbox2_area = (bbox2[2] - bbox2[0]) * (bbox2[3] - bbox2[1])
+
+            iou = inter_area / float(bbox1_area + bbox2_area - inter_area) if (bbox1_area + bbox2_area - inter_area) > 0 else 0
+            return iou
+
+        image_path = "./app/data1/" + ii
+        csv_file_path = output_csv_path = './app/Folder2/' + ii + 'bounding_boxes_with_recognition.csv'
+        image = cv2.imread(image_path)
+        font = cv2.FONT_HERSHEY_SIMPLEX
+        font_scale = 1.3
+        font_thickness = 2
+        color = (255, 0, 255)
+        bboxes = []
+        recognized_texts = []
+        with open(csv_file_path, 'r', encoding='utf-8') as csvfile:
+            csv_reader = csv.DictReader(csvfile)
+            for row in csv_reader:
+                bbox_match = re.search(r'\((\d+), (\d+), (\d+), (\d+)\)', row['Bounding Box'])
+                if bbox_match:
+                    bbox = [int(bbox_match.group(i)) for i in range(1, 5)]
+                    bboxes.append(bbox)
+                    recognized_texts.append(row['Recognized Text'])
+        filtered_bboxes = []
+        filtered_texts = []
+        iou_threshold = 0.4
+        for i, bbox1 in enumerate(bboxes):
+            keep = True
+            for j, bbox2 in enumerate(filtered_bboxes):
+                if calculate_iou(bbox1, bbox2) > iou_threshold:
+                    keep = False
+                    break
+            if keep:
+                filtered_bboxes.append(bbox1)
+                filtered_texts.append(recognized_texts[i])
+        for bbox, recognized_text in zip(filtered_bboxes, filtered_texts):
+            x1, y1, x2, y2 = bbox
+            cv2.rectangle(image, (x1, y1), (x2, y2), color, 2)
+            max_chars_per_line = 60
+            wrapped_text = textwrap.wrap(recognized_text, width=max_chars_per_line)
+            text_y = y1 - 10 if y1 - 10 > 10 else y1 + 10
+            for line in wrapped_text:
+                cv2.putText(image, line, (x1, text_y), font, font_scale, color, font_thickness)
+                text_y += int(font_scale * 20)
+            output_image_path = "./app/Folder3/" + "annotated" + ii + ".png"
+            cv2.imwrite(output_image_path, image)
+            print(f"Annotated image saved at {output_image_path}")
+            counter += 1
+        return output_image_path, output_csv_path,image_path
+
+@app.route('/download_csv/<filename>')
+def download_csv(filename):
+    return send_from_directory('./app/Folder2', filename, as_attachment=True)
+
+@app.route('/download_image/<filename>')
+def download_image(filename):
+    return send_from_directory('./app/Folder3', filename, as_attachment=True)
+
+@app.route('/uploads/<filename>')
+def serve_uploaded_file(filename):
+    return send_from_directory('./app/data1', filename)

requirements.txt

@@ -0,0 +1,14 @@
+flask
+pytesseract
+gunicorn
+easyocr
+psutil
+matplotlib
+tqdm
+requests
+pandas
+huggingface_hub
+supervision
+py-cpuinfo
+
+