Deep neural network processing for sensor blindness detection in autonomous machine applications

A method comprising:

• receiving image data representative of an image of a field of view of an image sensor of a vehicle in an environment;
• generating first data corresponding to annotations of regions of the image with compromised visibility;
• generating second data corresponding to one or more blindness classifications for each region of the regions of the image;
• generating third data corresponding to one or more blindness attributes for each blindness classification of the one or more blindness classifications;
• and
• training, using the first data, the second data, and the third data as ground truth data, a neural network to predict pixels of images with the compromised visibility, a blindness classification for each of the pixels, and a blindness attribute corresponding to the blindness classification.

A system comprising:

• one or more image sensors to generate image data of an environment around a vehicle;
• a computing device including one or more processing devices and one or more memory devices communicatively coupled to the one or more processing devices storing programmed instructions thereon, which when executed by the processor causes the instantiation of:
• a blindness determiner to:
• apply, to a neural network, the image data representative of an image;
• and
• compute, using the neural network and based at least in part on the image data, pixels of the image with compromised visibility;
• and
• a control component to perform one or more operations using the image based at least in part on the compromised visibility.