Select and manipulate Region of interest.
A small python module to select a polygonal region of interest (ROI) in a given image that is stored as a Shape object. You can use this Shape object later to manipulate the polygon selected. You can also extract the inner content from an image, calculate the histogram of the created shape, calculate the center of the shape, rotate the shape around its center, or translate the shape.
🚀 NEW: JAX-Enhanced Version - Now powered by JAX numpy for improved performance and parallel processing capabilities!
import cv2 as cv
from polyroi import Shapeimg = cv.imread('image.jpg')
shape = Shape.get_roi(img) #returns a Shape object
shape.draw_shape(img, color=(0, 255, 255), thickness=1)
while(1):
cv.imshow("Getting Started", img)
k = cv.waitKey(1) & 0xFF
if k == 27:
break
cv.destroyAllWindows()Process multiple ROI shapes simultaneously for improved performance:
from polyroi import Shape
import jax.numpy as jnp
# Create multiple shapes
shapes = [Shape([(0,0), (100,0), (100,100), (0,100)]) for _ in range(10)]
# Rotate all shapes in parallel
rotated_shapes = Shape.process_multiple_shapes_parallel(shapes, 'rotate', jnp.pi/4)
# Translate all shapes in parallel
translated_shapes = Shape.process_multiple_shapes_parallel(shapes, 'translate', 50, 30)Leverage JAX's vectorized operations for advanced processing:
# Convert shape to JAX array for advanced operations
shape_array = shape.to_array() # Returns JAX array
centroid = jnp.mean(shape_array, axis=0)
distances = jnp.linalg.norm(shape_array - centroid, axis=1)All mathematical operations now use JAX numpy, enabling automatic GPU acceleration when available.
Check out the examples/ directory for comprehensive demonstrations:
basic_jax_example.py- Basic JAX-enhanced operationsparallel_processing_example.py- Parallel batch processingimage_processing_example.py- Real-world image processing scenarios
pip install cv2
pip install numpy
pip install jax # New: For enhanced performance and GPU accelerationpip install polyroi Some time ago, I looked for an efficient tool to draw and manipulate polygons in a python environment. But I didn't find anything useful for my case. I did find some tools that can draw and extract a NumPy array, but as for the manipulation of shapes, I had to develop the logic myself. So I decided to create one. I was trying to implement the particle filter from Part-Based Lumbar Vertebrae Tracking in Videofluoroscopy Using Particle Filter. You can check the repository of how I did manage to work with this package.
img = cv.imread('image.jpg')
# returns a Shape object
shape = Shape.get_roi(img)
# Copy the shape
shape2 = Shape.copy(shape)
# Rotate the shape
shape2.rotate_around_center(np.pi/4)
# x translate the shape by 5
shape2.translate_x(5)
# y translate the shape by 5
shape2.translate_y(5)
# recalculate the center of the shape
shape2.centroid()
# translate the shape first point to (10, 15) along with the shape
shape2.translate_to(10, 15)
# x translate, y translate, and rotate around the center by np.pi / 12
shape2.update(5, 3, np.pi / 12)
# Drawing the shapes
shape.draw_shape(img, color=(0, 255, 255), thickness=1)
shape2.draw_shape(img, color=(0, 255, 0), thickness=1)
# return the bounding box points (upper left, bottom right)
p1, p2 = shape2.to_rectangle()
# plotting the image
while(1):
cv.imshow("Getting Started", img)
k = cv.waitKey(1) & 0xFF
if k == 27:
break
cv.destroyAllWindows()- @skywolfmo - Idea & Initial work
See also the list of contributors who participated in this project.