vprashant/doclayout_detector

Document Layout Detection

This script demonstrates how to use the document layout detection model on an image.
Below are the steps and code implementation.

---
## Step 1: Import Required Libraries

import cv2
import matplotlib.pyplot as plt
import numpy as np
from ultralytics import YOLO
from google.colab.patches import cv2_imshow

• cv2: For image processing.
• matplotlib.pyplot: For plotting if needed.
• numpy: For numerical operations.
• YOLOv10: For object detection.
• cv2_imshow: For displaying images in Google Colab.

---

Step 2: Load YOLOv10 Model

model = YOLO('vprashant/doclayout_detector/weights')  

• Load the YOLOv10 model with the path to your trained weights.

---

Step 3: Read and Prepare the Image

img = cv2.imread('/content/yolov10/dataset/train/images/11.png')
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

• Read the image from the specified path.
• Convert the image from BGR to RGB color space.

---

Step 4: Perform Object Detection

results = model(img)

• Run the YOLOv10 model on the image to get detection results.

---

Step 5: Extract and Process Detection Results

results = results[0] 
boxes = results.boxes
data = boxes.data.cpu().numpy()  

• Extract the first result (image-level result).
• Access the detected bounding boxes.
• Convert detection data to a NumPy array for processing.

---

Step 6: Visualize Results

for i, detection in enumerate(data):
    x1, y1, x2, y2, conf, cls_id = detection
    x1, y1, x2, y2 = map(int, [x1, y1, x2, y2])  # Convert coordinates to integers
    cv2.rectangle(img, (x1, y1), (x2, y2), (0, 255, 0), 2)  # Draw bounding box
    
    class_name = model.names[int(cls_id)]  # Get class name
    label = f"{class_name}: {conf:.2f}"  # Create label with confidence score
    cv2.putText(img, label, (x1, y1 - 10), cv2.FONT_HERSHEY_SIMPLEX, 
                0.9, (0, 255, 0), 2)  # Add text label

• Loop through all detections.
• Draw bounding boxes and labels on the image.

---

Step 7: Display the Processed Image

cv2_imshow(img) 
cv2.waitKey(0)
cv2.destroyAllWindows()

• Display the image with detections in Google Colab using cv2_imshow.
• Wait for a keypress and close any OpenCV windows.

---

Note

• Ensure you have the trained YOLOv10 model and the dataset in the specified paths.
• Replace the paths with your local or Colab paths.
• Install necessary libraries like OpenCV, Matplotlib, and ultralytics if not already installed.