| CPC G06V 20/584 (2022.01) [B60W 60/001 (2020.02); B60W 2420/403 (2013.01)] | 22 Claims |
|
1. A computer-implemented method implemented by one or more hardware processors, the computer-implemented method comprising:
obtaining, using the one or more hardware processors, data representing partial blockage images, each partial blockage image depicting blockage in front of a chroma key background;
generating, using the one or more hardware processors, a training data set including a plurality of synthetic partial blockage images and a plurality of non-synthetic partial blockage images, wherein each non-synthetic partial blockage image of the plurality of non-synthetic partial blockage images is associated with a low granularity label corresponding to a partial blockage, and wherein generating the training data set comprises, for each synthetic partial blockage image of the plurality of synthetic partial blockage images:
performing a chroma keying operation to extract imagery of a blockage from a partial blockage image of the partial blockage images;
superimposing the extracted imagery of the blockage on a background image to generate the synthetic partial blockage image; and
generating annotation data for the synthetic partial blockage image, the annotation data associated with a high granularity label distinguishing portions of the synthetic partial blockage image representing the extracted imagery of the blockage from portions of the synthetic partial blockage image representing the background image, wherein the high granularity label designates a blockage with respect to a more specific portion of an image than a low granularity label; and
training, using the one or more hardware processors, a neural network using the training data set to identify portions of input image data, corresponding to data from a sensor of an autonomous vehicle, that represent blockage on the sensor of the autonomous vehicle, wherein the neural network is trained to identify portions of input image data with high granularity, and wherein training comprises:
iteratively passing items from the training data set through the neural network to generate output with a high granularity label; and
updating the neural network based on a comparison of a label associated with an item from the training data set with:
for synthetic partial blockage images, the output of the neural network, and
for non-synthetic partial blockage images, a converted output of the neural network generated by converting the output from a high granularity to a low granularity.
|