Regarding MULTI LABEL CLASSIFICATION C2_W2_Classification with multiple outputs

How sigmoid Activation is able to predict car, bus, or pedestrian ?
Sigmoid outputs either 1 or 0 which basically means the probability of given input maybe
interpreted as
1 represents image contains car but not bus nor pedestrian or Vice - Versa .
But if image contains car and bus and no pedestrian , how to the sigmoid function is able to
differentiate car and bus if 1 represents the probability of image containing car and bus only.

Isn’t SOFTMAX function is a better option !

Each output unit detects one label.

And softmax applies to multi-class problem, not to multi-label problems.