LightGlue

The LightGlue model computes point matches between two images using SuperPoint and LightGlue.

class LightGlue()

use_local · boolean · default: "False"

If True, inference call is run on the local VM, else offloaded onto GRID-Cortex. Defaults to False.

def run()

image0 · np.ndarray · required

Input RGB image 1 of shape $(M,N,3)$.

image1 · np.ndarray · required

Input RGB image 2 of shape $(M,N,3)$.

Returns · List

List of point correspondences (matches) between the images.

points0 · np.ndarray

Feature points detected in image 1 of shape $(K,2)$.

points1 · np.ndarray

Feature points detected in image 2 of shape $(K,2)$.

Inference Call

from grid.model.perception.matching.glightglue import LightGlue
car = AirGenCar()

# We will be capturing an image from the AirGen simulator 
# and run model inference on it.

img =  car.getImage("front_center", "rgb").data
img1 = img.copy()

model = LightGlue(use_local = False)
matches1, points0_1, points1_1  = model.run(img, img1)

License

This code is licensed under the Apache 2.0 License.

Source

https://github.com/cvg/LightGlue