Abstract: Example implementations include a method, apparatus and computer-readable medium for estimating a real world distance in an image, comprising obtaining an image from a camera installed in an environment and identifying, in the image, a first set of parallel lines on a real world ground plane and a second set of parallel lines on the real world ground plane. The implementations include determining a first intersection point of the first set of parallel lines and a second intersection point of the second set of parallel lines, identifying a ground line connecting the first intersection point and the second intersection point, and estimating an internal camera matrix based on a focal length and an image size associated with the camera. The implementations include calibrating the camera for real world distance measurement by determining a perpendicular vector to the ground line based on the internal camera matrix.

Abstract: A security device maintenance system is provided. The aspects include one or more processors configured to assess security data quality by conducting one or more tests that compare the security data to one or more predetermined thresholds, assign one or more anomaly indicators to security devices if the security data quality fails to meet the one or more predetermined thresholds, and generate an automated maintenance report for the one or more security devices including the one or more anomaly indicators. The aspects can also include periodically retrieving one or more security data from the one or more security devices, utilizing an artificial intelligence (AI) model to detect one or more anomalies, and generating one or more predictive maintenance suggestions.

Abstract: A system may be configured to provide natural language search over security videos. In some aspects, the system may generate a first representation of sampled video information in a multidimensional format via a first machine learning model, and receive a request including a natural language input. Further, the system may generate a second representation of the natural language input in the multidimensional format via a second machine learning model that is a different from the first machine learning model, and determine that the first representation has a predefined relationship with the second representation. In addition, the system may present the second representation as a response to the request based on the first representation having the predefined relationship with the second representation.

Abstract: Systems and methods are disclosed for indicating occupancy of a point of interest in a room. A video feed from a camera located in the room can be received, and a person in the video feed can be detected by using an artificial intelligence (AI) model. A location of the person on a floor plan of the room can be identified based on a video feed location of the person detected in the video feed, and where the location of the person on the floor plan is at the point of interest for at least a threshold period of time, the point of interest can be indicated as being occupied.

Abstract: Examples relate to processing a first image captured at a first time to determine a first set of multiple key points of a face in the first image, processing a second image captured at a second time to determine a second set of the multiple key points on the face in the second image, determining a location of at least a portion of the multiple key points in a three-dimensional space based on a first location of each of the portion of the multiple key points in the first image and a second location of each of the portion of the multiple key points in the second image, and detecting whether the face is a valid three-dimensional face for facial recognition based at least in part on the location of at least the portion of the multiple key points in the three-dimensional space.

Abstract: Example implementations include a method, apparatus, and computer-readable medium for object detection, comprising detecting a first object in a first image frame and a second image frame, wherein the first object is bounded by region-of-interest (ROI) boundaries generated by an ROI detection model. The implementations further include calculating a speed of the first object using positions of the first object in the first and second image frame, identifying at least one image frame that should include the first object based on a calculated speed of the first object. The implementations further include determining that the at least one image frame should be added to a training dataset for the ROI detection model in response to detecting that the ROI detection model did not generate a ROI boundary in the at least one image frame, and subsequently re-training the ROI detection model using said training dataset.

Abstract: A video camera recording system is provided. The aspects include one or more memories configured to store program code. The program code is for performing an automated verification test of elements of the video camera recording system using at least one artificial intelligence (AI) model to identify potential issues with the elements of the video camera recording system. The aspects further include one or more processors, operatively coupled to the one or more memories and configured to run the program code. The aspects also include a transceiver configured to transmit instructions causing a corrective action to be performed for at least one of the potential issues.

Abstract: Examples relate to processing a first image captured by a camera at a first time to determine a first set of multiple key points of a face in the first image, processing a second image captured by the camera at a second time to determine a second set of the multiple key points on the face in the second image, determining a location of at least a portion of the multiple key points in a three-dimensional space based on a first location of each of the portion of the multiple key points in the first image and a second location of each of the portion of the multiple key points in the second image, and detecting whether the face is a valid three-dimensional face for facial recognition based at least in part on the location of at least the portion of the multiple key points in the three-dimensional space.

Abstract: Disclosed herein is an object detection system, including apparatuses and methods for object detection. An implementation may include receiving a first image frame from an ROI detection model that generated a first ROI boundary around a first object detected in the first image frame and subsequently receiving a second image frame. The implementation further includes predicting, using an ROI tracking model, that the first ROI boundary will be present in the second image frame and then detecting whether the first ROI boundary is in fact present in the second image frame. The implementation includes determining that the second image frame should be added to a training dataset for the ROI detection model when detecting that the ROI detection model did not generate the first ROI boundary in the second image frame as predicted and re-training the ROI detection model using the training dataset.

Abstract: Systems and methods are disclosed relating to autonomous building security recommendation generation. For example, a method can include receiving, by one or more processors, sensor data from one or more sensors associated with a building system. The method can further include determining, by the one or more processors using a machine learning model and the sensor data, a recommended action for an operator to perform, the machine learning model trained using training data comprising data retrieved from one or more data sources maintained by at least one of a first entity associated with the building system or a second entity associated with the one or more sensors. The method can further include presenting, by the one or more processors using at least one of a display device or an audio output device, a notification corresponding to the recommended action.

Abstract: Example implementations include a method, apparatus and computer-readable medium for detecting an object in an image, by applying, on a first input image, a first object detector configured to generate a bounding box around the object in the first input image. The implementations further include identifying a first bounding box having a first bounding box size around the object in the first input image and determining that the first bounding box size is less than a threshold size. Additionally, the implementations further include generating a second input image by cropping an area of the original image corresponding to the first bounding box and identifying, by applying a second object detector on the second input image, a second bounding box around the object in the second input image. Additionally, the implementations further include performing an image analysis function on the object using information from the second bounding box.

Abstract: Example implementations include a method, apparatus and computer-readable medium for estimating a real world distance in an image, comprising obtaining an image from a camera installed in an environment and identifying, in the image, a first set of parallel lines on a real world ground plane and a second set of parallel lines on the real world ground plane. The implementations include determining a first intersection point of the first set of parallel lines and a second intersection point of the second set of parallel lines, identifying a ground line connecting the first intersection point and the second intersection point, and estimating an internal camera matrix based on a focal length and an image size associated with the camera. The implementations include calibrating the camera for real world distance measurement by determining a perpendicular vector to the ground line based on the internal camera matrix.

Abstract: Example implementations include a method, apparatus and computer-readable medium for estimating a real world height in an image, comprising obtaining two images of an object, with known height, wherein each image depicts the object at a different position that is estimated using a homography matrix. The homography matrix transforms an image plane associated with the image to a real world plane. The implementations include generating a first vector that intersects with the object and points to a real world ground plane using the first image and a second vector that intersects with the same point on the object and points to the real world ground plane using the second image. Additionally, the implementations further include calibrating the camera by determining a height of the center of a camera lens of the camera using an intersection of the first vector and the second vector.

Abstract: Disclosed herein is an object detection system, including apparatuses and methods for object detection. An implementation may include receiving a first class of a first object depicted in an image frame from a classification model and subsequently receiving a second image frame. The implementation further includes predicting, using a classification tracking model, that the classification model will output the first class for the second image frame and then detecting whether the first class is in fact outputted. The implementation includes determining that the second image frame should be added to a training dataset for the classification model when detecting that the classification model did not generate the first class for the second image frame as predicted and re-training the classification model using the training dataset.

Abstract: Disclosed herein is an object detection system, including apparatuses and methods for object detection. An implementation may include receiving a first image frame from an ROI detection model that generated a first ROI boundary around a first object detected in the first image frame and subsequently receiving a second image frame. The implementation further includes predicting, using an ROI tracking model, that the first ROI boundary will be present in the second image frame and then detecting whether the first ROI boundary is in fact present in the second image frame. The implementation includes determining that the second image frame should be added to a training dataset for the ROI detection model when detecting that the ROI detection model did not generate the first ROI boundary in the second image frame as predicted and re-training the ROI detection model using the training dataset.

Abstract: Disclosed herein is an object detection system, including apparatuses and methods for object detection. An implementation may include receiving a first image frame from an ROI detection model that generated a first ROI boundary around a first object detected in the first image frame and subsequently receiving a second image frame. The implementation further includes predicting, using an ROI tracking model, that the first ROI boundary will be present in the second image frame and then detecting whether the first ROI boundary is in fact present in the second image frame. The implementation includes determining that the second image frame should be added to a training dataset for the ROI detection model when detecting that the ROI detection model did not generate the first ROI boundary in the second image frame as predicted and re-training the ROI detection model using the training dataset.

Abstract: Examples relate to processing a first image captured by a camera at a first time to determine a first set of multiple key points of a face in the first image, processing a second image captured by the camera at a second time to determine a second set of the multiple key points on the face in the second image, determining a location of at least a portion of the multiple key points in a three-dimensional space based on a first location of each of the portion of the multiple key points in the first image and a second location of each of the portion of the multiple key points in the second image, and detecting whether the face is a valid three-dimensional face for facial recognition based at least in part on the location of at least the portion of the multiple key points in the three-dimensional space.

Abstract: Examples described herein generally relate to processing a first image captured by the first camera at a first time to determine a first set of multiple key points of a face in the first image, processing a second image captured by the second camera at a second time, that is equal to or within a threshold time of the first time, to determine a second set of the multiple key points on the face in the second image, determining a location of at least a portion of the multiple key points in a three-dimensional space based on a first location of each of the portion of the multiple key points in the first image and a second location of each of the portion of the multiple key points in the second image, and detecting whether the face is a valid three-dimensional face for facial recognition based at least in part on the location of at least the portion of the multiple key points in the three-dimensional space.

Abstract: A building radar-camera system includes a camera configured to capture one or images, the one or more images including first locations within the one or more images of one or more points on a world-plane and a radar system configured to capture radar data indicating second locations on the world-plane of the one or more points. The system includes one or more processing circuits configured to receive a correspondence between the first locations and the second locations of the one or more points, generate a sphere-to-plane homography, the sphere-to-plane homography translating between points captured by the camera modeled on a unit-sphere and the world-plane based on the correspondence between the first locations and the second locations, and translate one or more additional points captured by the camera or captured by the radar system between the unit-sphere and the world-plane based on the sphere-to-plane homography.

Abstract: Disclosed herein is an object detection system, including apparatuses and methods for object detection. An implementation may include receiving a first image frame from an ROI detection model that generated a first ROI boundary around a first object detected in the first image frame and subsequently receiving a second image frame. The implementation further includes predicting, using an ROI tracking model, that the first ROI boundary will be present in the second image frame and then detecting whether the first ROI boundary is in fact present in the second image frame. The implementation includes determining that the second image frame should be added to a training dataset for the ROI detection model when detecting that the ROI detection model did not generate the first ROI boundary in the second image frame as predicted and re-training the ROI detection model using the training dataset.