SAM Prompt Encoder

Problem

Build sparse prompt embeddings for the Segment Anything prompt encoder. Each point contributes [x, y, is_positive, is_background], where the label 1 sets is_positive and the label 0 sets is_background.

The signature is sam_prompt_encoder(points, labels). Return the list of embeddings.

Hint

Positive clicks select an object while negative clicks exclude regions.

Test cases

case 1

in: [[[0.5,0.5]],[1]]

expected: [[0.5,0.5,1,0]]

case 2

in: [[[0.1,0.2]],[0]]

expected: [[0.1,0.2,0,1]]

case 3

in: [[[-0.5,1.5],[0.3,0.3]],[1,0]]

expected: [[-0.5,1.5,1,0],[0.3,0.3,0,1]]

case 4

in: [[[0,0]],[1]]

expected: [[0,0,1,0]]

Starter codePython

def sam_prompt_encoder(points, labels):
    # Your code here
    pass

4 test cases · run your code against them in the editor