Abstract: The invention provides the techniques and systems that allow for the identification and classification of objects (i.e., humans) in images using a predictive segmentation model. More specifically, human forms are identified within an image by generating pixel-level bounding boxes for each possible object and using offsets and segmentation masking. In some instances, embodiments of the invention use an identified floor plane that intersects with a bounding box to identify a three-dimensional position for the intersection point, which can then be assigned to the human form and represent its depth within the image.

Abstract: The invention provides the techniques and systems that allow for an accurate positioning of a human in a virtual or augmented reality scene based on monocular video of an object without the need for multiple cameras, specialized equipment or robotic models of the human anatomy. The object is modeled by assigning reference points to elements of the object and identifying anchor points and initial reference points. Based on two dimensional coordinates of the anchor points, two dimensional coordinates of the anchor points not contained in the subset of initial reference points are calculated, and based on the anchor points, a depth value is calculated for each anchor point, and three dimensional position values are assigned to each of the anchor points. Based on the three dimensional position value of an adjacent reference point, a three dimensional position value is iteratively assigned to each of the remaining reference points.

Abstract: Methods and supporting systems calculate a three-dimensional orientation of a camera based on images of objects within the cameras field of view. A camera (such as a camera within a mobile phone), captures two-dimensional video content including a human body and at least one additional object and assigns a frame of the video content at a first time as an anchor frame. The human body within the anchor frame is modeled by assigning points of movement to a set of body elements. A subsequent frame of the video content is received in which the human body has moved, and a translation function is derived that calculates the three-dimensional position of each of the body elements based on two dimensional movements of the body elements between the anchor frame and the subsequent frame. Using the translation function, a three-dimensional relationship of the camera and the additional object is calculated.