Shared median-scaling metric for multi-camera self-supervised depth evaluation

A non-transitory computer-readable medium having program code recorded thereon for multi-camera self-supervised depth evaluation, the program code being executed by a processor and comprising:

• program code to train a self-supervised depth estimation model and an ego-motion estimation model according to a multi-camera photometric loss associated with a multi-camera rig of an ego vehicle;
• program code to generate a single-scale correction factor according to a depth map of each camera of the multi-camera rig during a time-step;
• program code to predict a 360° point cloud of a scene surrounding the ego vehicle according to the self-supervised depth estimation model and the ego-motion estimation model;
• and
• program code to scale the 360° point cloud according to the single-scale correction factor to form an aligned 360° point cloud.

A system for multi-camera self-supervised depth evaluation, the system comprising:

• a self-supervised depth estimation model and an ego-motion estimation model trained according to a multi-camera photometric loss associated with a multi-camera rig of an ego vehicle;
• a single-scale correction factor module to generate a single-scale correction factor according to a depth map of each camera of the multi-camera rig during a time-step;
• a full surround mono-depth (FSM) point cloud module to predict a 360° point cloud of a scene surrounding the ego vehicle according to the self-supervised depth estimation model and the ego-motion estimation model;
• and
• a point cloud scaling module to scale the 360° point cloud according to the single-scale correction factor to form an aligned 360° point cloud.