Abstract: In some aspects, there is provided a method for determining a rotational orientation of an object in an image. The image depicts the object in a scene. The method includes providing images depicting the object in the scene to a trained statistical model. The images depict the scene of the image at different rotation angles. A rotation angle of a respective image corresponds to a potential rotational orientation of the object depicted in the respective image. The method further includes, in response to the providing, receiving, for each image, a confidence score indicating a likelihood generated by the trained statistical model that the object is at the potential rotational orientation corresponding to the rotation angle of the respective image. The method further includes determining the rotational orientation of the object in the image based at least in part on an analysis of the confidence scores and respective potential rotational orientations.

Abstract: In some aspects, there is provided a method for determining a rotational orientation of an object in an image. The image depicts the object in a scene. The method includes providing images depicting the object in the scene to a trained statistical model. The images depict the scene of the image at different rotation angles. A rotation angle of a respective image corresponds to a potential rotational orientation of the object depicted in the respective image. The method further includes, in response to the providing, receiving, for each image, a confidence score indicating a likelihood generated by the trained statistical model that the object is at the potential rotational orientation corresponding to the rotation angle of the respective image. The method further includes determining the rotational orientation of the object in the image based at least in part on an analysis of the confidence scores and respective potential rotational orientations.