petyr

Affine, Similarity Transformations and Homography for Python. Fast and chainable inplace operations. Produced matrix can be used with cv2.warpAffine and cv2.warpPerspective.

Installation

For stable release

pip3 install petyr

Or, build from source

pip3 install git+https://github.com/safwankdb/petyr

Usage

from petyr import Similarity, Affine, Homography

Affine, Similarity and Homography derive from the Transform2D base class which implements all the basic operations.

Applying Transformation

p = np.array([[0,0],[1,0],[1,1],[0,1]])

rotate_and_move = Affine().rotate(90).translate(2,1)
q = rotate_and_move * p

pt = Homography.from_elements([1,0,0,0,1,0,0.1,0.2,1])
r = pt * p

Finding Transformation

at = Affine.from_points(p, q)
pt = Homography.from_points(p, q)
st = Similarity.from_points(p, q)

Basic Operations

These operations modify the object in-place except for invert() which return a new object.

• Translation
• Scaling
• Shearing
• Rotation
• Inversion
• Reset

at = Affine()
at.translate(1, 3)
at.scale(1.05, 2)
at.rotate(45, degrees=True)
at.shear(10, 45)
at_inv = at.invert()

Same goes for Homography and Transform2D objects.

Chaining

Chaining Operations

Mutiple operations can be chained together.

at = Affine()
at.scale(2,2).rotate(90)
at.shear(10, 0).translate(-3, 4)

Chaining Transforms

Multiple transforms can be multiplied together. This is a non-commutative operation. The rightmost transform will be applied first.

a = Affine()
a.translate(2,3)
b = Homography()
b.scale(4,5)
c = a * b

Testing

To run the inbuilt unit tests,

git clone https://github.com/safwankdb/petyr
cd petyr
python3 -m unittest -v

Maitainers

• Mohd Safwan

TODO

• Implement petyr.Homography.
• Add unit tests.
• Vectorize from_points.
• Add unit tests for Affine and Homography classes as well.
• Add Similarity class for similarity transforms.
• Update all tests for similarity transform
• Update README with Similarity Transform
• Generate complete documentation.
• Move documentation to somwhere other than README.