Segment Anything 2

The SAM2 class implements a wrapper for the Segment Anything 2.1 (SAM2) model, which segments objects in RGB images and videos based on input prompts.

boolean

default:"False"

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

np.ndarray

required

The input RGB image of shape

(M, N, 3)

np.ndarray

required

Points in the form of coordinates of shape

(x, y)

to segment.

np.ndarray

required

The labels of the points to segment.

boolean

default:"False"

If True, returns a list of masks of shape

(M,N)

. Defaults to False. The mutlimask feature is only available for local inference.

np.ndarray

The predicted output.

from grid.model.perception.segmentation.sam2 import SAM2
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
prompts = np.array([[616, 208]])
labels = np.array([1])

model = SAM2(use_local = False)
result = model.run(rgbimage=img, prompts, labels)
print(result.shape)

License
Source

This code is licensed under the Apache 2.0, and BSD-3 License.

from grid.model.perception.segmentation.sam2 import SAM2 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 prompts = np.array([[616, 208]]) labels = np.array([1]) model = SAM2(use_local = False) result = model.run(rgbimage=img, prompts, labels) print(result.shape)

Get Started

Open GRID

GRID Enterprise

Robot API

Simulation

AI Layer

Deployment

Data Generation Pipelines

FAQ