Abstract: A system for generating accurate training data for a machine model, and a system for training and using the machine model are disclosed. The training data is generated by taking video of a subject with two cameras, for example, an infrared (IR) camera and a visible light camera. Prior to taking the video, the subject is marked with an IR marker at key points. The key point marks are shown in the IR video and are not shown in the visible light video. The IR video is processed to generate an IR mark video emphasizing the IR marks. To train the ML model, the visible video is used as the input and the IR mark video is used as target information. Once trained, the machine learning model can generate key point marks based on video of subjects which are not marked.

Abstract: Provided are systems and methods for tracking device engagement The system includes a viewing device (e.g., television, etc.) and an eye-tracking device (e.g., a camera, etc.). The eye-tracking device is configured to capture data about the gaze of a viewer to determine if the viewer is watching the television at any given point in time. This information may be used for a variety of purposes, such as tracking user engagement with advertisement content. As another example, the information may be used for device energy saving purposes. For example, a screen of the device can be dimmed or turned off if the gaze of the viewer has not been directed towards the television for a given period of time. A notification may also be presented to the viewer prompting the viewer to indicate if they are still viewing the content being presented on the device.

Abstract: A computer-implemented method is disclosed. The method includes selecting one or more target surfaces portrayed in at least one video frame, generating a video data latent space representation of the at least one video frame, accessing a plurality of supplemental data latent space representations of a plurality of supplemental data sets, identifying a particular supplemental data latent space representation based at least in part on the video data latent space representation, selecting a particular supplemental data set in response to identifying the particular supplemental data latent space representation, the particular supplemental data set corresponding with the particular supplemental data latent space representation, and inserting the particular supplemental data set into the at least one video frame.