Abstract: Velocity adjustments based on roadway scene comprehension may include controlling an ego vehicle traveling in a first lane, wherein the first lane comprises a lane of travel for the ego vehicle of a multilane roadway including at least the first lane and a second lane; capturing object ranges by the ego vehicle using at least a first sensor corresponding to a first sensor type; identifying a first vehicle operating in the second lane from the first sensor; determining a first velocity of the first vehicle; generating a first lane flow rate for the second lane from one or more captured object velocities, wherein the lane flow rate of the second lane is based on at least the first velocity of the first vehicle and a traffic density estimation; determining whether to modify an ego vehicle velocity based on the first lane flow rate; and in response to determining whether to modify the ego vehicle velocity, generating instructions for the ego vehicle's velocity.

Abstract: Dynamic calibration of cameras in a stereoscopic configuration may include: determining a disparity between first image data from a first camera and second image data from a second camera, wherein the first camera and the second camera are in a stereoscopic configuration, and wherein the disparity comprises a difference in placement of one or more objects in the first image data relative to the second image data; and adjusting one or more of the first camera or the second camera, based on the disparity and sensor data from a sensor other than the first camera and the second camera, to calibrate the stereoscopic configuration of the first camera and the second camera to achieve stereoscopic camera distance functionality.

Abstract: The present disclosure provides an improved training methodology that enables supervised contrastive learning to be simultaneously performed across multiple positive and negative training examples. In particular, example aspects of the present disclosure are directed to an improved, supervised version of the batch contrastive loss, which has been shown to be very effective at learning powerful representations in the self-supervised setting Thus, the proposed techniques adapt contrastive learning to the fully supervised setting and also enable learning to occur simultaneously across multiple positive examples.

Abstract: The present disclosure provides an improved training methodology that enables supervised contrastive learning to be simultaneously performed across multiple positive and negative training examples. In particular, example aspects of the present disclosure are directed to an improved, supervised version of the batch contrastive loss, which has been shown to be very effective at learning powerful representations in the self-supervised setting. Thus, the proposed techniques adapt contrastive learning to the fully supervised setting and also enable learning to occur simultaneously across multiple positive examples.

Abstract: The present disclosure provides an improved training methodology that enables supervised contrastive learning to be simultaneously performed across multiple positive and negative training examples. In particular, example aspects of the present disclosure are directed to an improved, supervised version of the batch contrastive loss, which has been shown to be very effective at learning powerful representations in the self-supervised setting. Thus, the proposed techniques adapt contrastive learning to the fully supervised setting and also enable learning to occur simultaneously across multiple positive examples.