Abstract: Embodiments described herein provide systems and methods for training video models to perform a task from an input instructional video. A procedure knowledge graph (PKG) may be generated with nodes representing procedure steps, and edges representing relationships between the steps. The PKG may be generated based on text and/or video training data which includes procedures (e.g., instructional videos). Using the PKG, a video model may be trained using the PKG to provide supervisory training signals for a number of tasks. Once the model is trained, it may be fine-tuned for a specific task which benefits from the model being trained in a way that makes the model embed procedural information when encoding videos.

Abstract: Methods and systems for time marking of media items at a platform using machine-learning are provided herein. An indication of a identified media item is provided as input to a machine-learning model and one or more outputs of the machine-learning model is obtained. The one or more obtained outputs comprise time marks identifying each of the plurality of content segments of the media item. Each of the plurality of content segments is associated with a segment start indicator for a timeline of the media item. A resulting duration is determined of a combination of the plurality of content segments for which the time marks were obtained from the one or more of outputs of the machine-learning model. Responsive to determining that the resulting duration is less than the duration of the media item, one or more further inputs is provided to the machine learning model.

Abstract: Methods and systems for time marking of media items at a platform using machine-learning are provided herein. A media item is provided to users of a platform. An indication of the identified media item is provided as input to a machine-learning model that is trained using different feature types of historical media items to predict a plurality of content segments of a given media item each depicting, to the one or more users, a distinct section of the media item. One or more outputs of the machine-learning model are obtained comprising time marks identifying each of the plurality of content segments of the media item. Each of the plurality of content segments are associated with a segment start indicator for a timeline of the media item. The media item and an indication of each segment start indicator is provided for presentation to at least one user.

Abstract: A method for using a multi-hop reasoning framework to perform multi-step compositional long-term reasoning is presented. The method includes extracting feature maps and frame-level representations from a video stream by using a convolutional neural network (CNN), performing object representation learning and detection, linking objects through time via tracking to generate object tracks and image feature tracks, feeding the object tracks and the image feature tracks to a multi-hop transformer that hops over frames in the video stream while concurrently attending to one or more of the objects in the video stream until the multi-hop transformer arrives at a correct answer, and employing video representation learning and recognition from the objects and image context to locate a target object within the video stream.

Abstract: A method for compositional reasoning of group activity in videos with keypoint-only modality is presented. The method includes obtaining video frames from a video stream received from a plurality of video image capturing devices, extracting keypoints all of persons detected in the video frames to define keypoint data, tokenizing the keypoint data with time and segment information, clustering groups of keypoint persons in the video frames and passing the clustering groups through multi-scale prediction, and performing a prediction to provide a group activity prediction of a scene in the video frames.

Abstract: A method for using a multi-hop reasoning framework to perform multi-step compositional long-term reasoning is presented. The method includes extracting feature maps and frame-level representations from a video stream by using a convolutional neural network (CNN), performing object representation learning and detection, linking objects through time via tracking to generate object tracks and image feature tracks, feeding the object tracks and the image feature tracks to a multi-hop transformer that hops over frames in the video stream while concurrently attending to one or more of the objects in the video stream until the multi-hop transformer arrives at a correct answer, and employing video representation learning and recognition from the objects and image context to locate a target object within the video stream.