Abstract: A video-processing technique uses machine-trained logic to detect and track people that appear in video information. The technique then ranks the prominence of these people in the video information, to produce ranking information. The prominence of a person reflects a level of importance of the person in the video information, corresponding to the capacity of the person to draw the attention of a viewer. For instance, the prominence of the person reflects, at least in part, an extent to which the person appears in the video information. The technique performs its ranking based on person-specific feature information. The technique produces each instance of person-specific feature information by accumulating features pertaining to a particular person. One or more application systems make use of the ranking information to control the presentation of the video information.

Abstract: Examples of the present disclosure describe systems and methods for video genre classification. In one example implementation, video content is received. A plurality of sliding windows of the video content is sampled. The plurality of sliding windows comprises audio data and video data. The audio data is analyzed to identify a set of audio features. The video data is analyzed to identify a set of video features. The set of audio features and the set of video features is provided to a classifier. The classifier is configured to detect a genre for the video content using the set of audio features and the set of video features. The video content is indexed based on the genre.

Abstract: Examples of the present disclosure describe systems and methods for on-device, in-browser AI processing. In examples, a selection of an AI pipeline is received. Content associated with the AI pipeline is also received. The content is segmented into multiple data segments and a set of data features is generated for the data segments. AI modules associated with the AI pipeline are loaded to create the AI pipeline. The set of data features is provided to the AI pipeline. The AI pipeline is executed to generate insights for the set of data features. The insights are then provided to a user.

Abstract: Examples of the present disclosure describe improved systems and methods for detecting keywords in audio content. In one example implementation, audio content is segmented into one or more audio segments. One or more text segments is generated, each text segment corresponding to each of the audio segments. For each text segment, one or more phrase candidate values is generated using a textual analysis, and one or more sentence embedding values is generated using a sentence embedding analysis. Next, an average sentence embedding value is calculated using the one or more sentence embedding values. Each of the one or more phrase candidate values is compared to the average sentence embedding value. Each phrase candidate value having a comparison value above a threshold value is labeled as representing a keyword.

Abstract: Systems, methods, and a computer-readable medium are provided for matching textless elements to texted elements in video content. A video processing system including a textless matching system may divide a video into shots, identify shots having similar durations, identify sequences of shots having similar durations, and compare image content in representative frames of the sequences to determine whether the sequences match. When the sequences are determined to match, the sequences may be paired, wherein the first sequence may include shots with overlaid text and the second sequence may include textless version of corresponding texted shots included in the first sequence. In some examples, the video processing system may further replace the determined corresponding texted shots.

Abstract: A video indexing system identifies groups of frames within a video frame sequence captured by a static camera during a same scene. Context metadata is generated for each frame in each group based on an analysis of fewer than all frames in the group. The frames are indexed in a database in association with the generated context metadata.

Abstract: A computer-implemented method of accounting for visibility of a first attribute of one or more attributes associable with an object presented in an image is provided. The method includes inputting a training image of a first object into an attribute identification machine learning model, the training image being associated with labeled visibility data indicating whether the first attribute is visible in the inputted training image, generating, based on the inputted training image, visibility prediction data representing a prediction by the attribute identification machine learning model as to whether the first attribute is predicted to be visible in the inputted training image, comparing the generated visibility prediction data with labeled visibility data, and modifying the attribute identification machine learning model based on the comparison of the generated visibility prediction data and the labeled visibility data.

Abstract: A video navigation and search tool includes a user interface that facilitates user interactions with indexed video content stored in a database. The user interface includes a library of thumbnail images that each individually depict a different subject associated in memory with a different detection identifier (ID). Each of the thumbnail images in the library is an image cropped from a single frame of a video file. Responsive to receiving a user selection of one of the thumbnail images associated with a first detection ID, the video navigation and search tool retrieves context metadata identifying frames in the video file indexed in the database in association with the first detection ID and presents video segment information on the user interface. The presented video segment information identifies one or more segments in the video file including the frames associated with the first detection ID.

Abstract: The present invention generates a compression invariant motion timeline for a video. For each second in the video, the method uses an identification procedure to determine a stored key frame size, a frame counter, and an accumulated size of frames. The method then calculates and stores a motion value for the second using the stored key frame size, the frame counter, and the accumulated size of frames. The motion values for each second may be used to construct a timeline for the video.

Abstract: An approach using 3D algorithms to solve 2D head localization problems is disclosed. A system can extrapolate aspects of one part of an object, e.g., extract characteristics of a person's head, using a 2D input image of another part of the object, e.g., a 2D image of the person's face. The system then selects an appropriate 3D model by the use of facial features detected in an image of a person's face. Using the selected 3D model and the 3D rotation angles provided by a face detector, the system rotates the model and then projects the model to a 2D shape. The system then scales and translates, e.g., transforms, the 2D shape to match the 2D face bounding box. Then, using the transformed 2D shape, the system extracts a bounding box for the extracted portion of an object, e.g., the head of the person depicted in the 2D input image.

Abstract: Tracking of targets in video captured by a multi-camera surveillance system is often difficult due to incomplete camera coverage and due to the complexities associated with automated recognition in a dense and highly variable environment. The present disclosure embraces a system and method for target tracking in a multi-camera surveillance system that reduces the time required for a user to track a target by offering computer generated search results that include suggestions of candidates found in the video that match (to some degree) the target. When one of the candidates is selected by a user, the search is shifted to other cameras and continued. In this way, the target may be tracked quickly. Further, a playlist of video snippets of the target is accumulated during the search and can be played to show all captured video of the target moving about the facility.

Abstract: Currently, systems for assessing traffic, such as retail traffic, only output counting results in numeric measures. Current systems do not assess problems and/or provide solutions to problems using traffic analysis methods or applications. The present disclosure is directed to methods, systems and media for applying retail traffic analysis statistics to provide actionable intelligence, such as solutions to particular inquiries or problems.

Abstract: An approach using 3D algorithms to solve 2D head localization problems is disclosed. A system can extrapolate aspects of one part of an object, e.g., extract characteristics of a person's head, using a 2D input image of another part of the object, e.g., a 2D image of the person's face. The system then selects an appropriate 3D model by the use of facial features detected in an image of a person's face. Using the selected 3D model and the 3D rotation angles provided by a face detector, the system rotates the model and then projects the model to a 2D shape. The system then scales and translates, e.g., transforms, the 2D shape to match the 2D face bounding box. Then, using the transformed 2D shape, the system extracts a bounding box for the extracted portion of an object, e.g., the head of the person depicted in the 2D input image.

Abstract: Tracking of targets in video captured by a multi-camera surveillance system is often difficult due to incomplete camera coverage and due to the complexities associated with automated recognition in a dense and highly variable environment. The present disclosure embraces a system and method for target tracking in a multi-camera surveillance system that reduces the time required for a user to track a target by offering computer generated search results that include suggestions of candidates found in the video that match (to some degree) the target. When one of the candidates is selected by a user, the search is shifted to other cameras and continued. In this way, the target may be tracked quickly. Further, a playlist of video snippets of the target is accumulated during the search and can be played to show all captured video of the target moving about the facility.

Abstract: A video indexing system identifies groups of frames within a video frame sequence captured by a static camera during a same scene. Context metadata is generated for each frame in each group based on an analysis of fewer than all frames in the group. The frames are indexed in a database in association with the generated context metadata.

Abstract: The present invention generates a compression invariant motion timeline for a video. For each second in the video, the method uses an identification procedure to determine a stored key frame size, a frame counter, and an accumulated size of frames. The method then calculates and stores a motion value for the second using the stored key frame size, the frame counter, and the accumulated size of frames. The motion values for each second may be used to construct a timeline for the video.

Abstract: A system and method to determine a flow direction automatically so that counter-flow motion may be detected without user calibration is disclosed. The system and method utilize information available from existing video compression algorithms to sense motion and determine a flow direction. Once a flow direction for an area is determined, the system and method can detect counter-flow motion using compressed video without any decompression, thereby improving the speed of detection and simplifying the computational complexity. These improvements facilitate counter-flow alarms and expand the information available to a user monitoring the video.

Abstract: The present invention generates a compression invariant motion timeline for a video. For each second in the video, the method uses an identification procedure to determine a stored key frame size, a frame counter, and an accumulated size of frames. The method then calculates and stores a motion value for the second using the stored key frame size, the frame counter, and the accumulated size of frames. The motion values for each second may be used to construct a timeline for the video.

Abstract: The present invention generates a compression invariant motion timeline for a video. For each second in the video, the method uses an identification procedure to determine a stored key frame size, a frame counter, and an accumulated size of frames. The method then calculates and stores a motion value for the second using the stored key frame size, the frame counter, and the accumulated size of frames. The motion values for each second may be used to construct a timeline for the video.

Abstract: A method of monitoring the amount of time spent in a specified area by an individual comprises employing a first camera to automatically create one or more entrance images, each entrance image containing a face of an entering individual that passes a first location, and storing each entrance image in a database along with a corresponding entrance time that the entering individual passed the entrance location. An exit image is compared to the entrance images in the database to identify a matching entrance image containing the same face as the exit image. A stay time is then determined for the exiting individual by determining the difference between the entrance time corresponding to the matching entrance image and the exit time.