Abstract: A device configured to receive a first encoded vector and receive one or more feature descriptors for a first object. The device is further configured to remove one or more encoded vectors from an encoded vector library that are not associated with the one or more feature descriptors and to identify a second encoded vector in the encoded vector library that most closely matches the first encoded vector based on the numerical values within the first encoded vector. The device is further configured to identify a first item identifier in the encoded vector library that is associated with the second encoded vector and to output the first item identifier.

Abstract: A system for capturing images for training an item identification model obtains an identifier of an item. The system detects a triggering event at a platform, where the triggering event corresponds to a user placing the item on a platform. The system causes the platform to rotate. The system causes at least one camera to capture an image of the item while the platform is rotating. The system extracts a set of features associated with the item from the image. The system associates the item to the identifier and the set of features. The system adds a new entry to a training dataset of the item identification model, where the new entry represents the item labeled with the identifier and the set of features.

Abstract: A device configured to capture a first overhead depth image of the platform using a three-dimensional (3D) sensor at a first time instance and a second overhead depth image of a first object using the 3D sensor at a second time instance. The device is further configured to determine that a first portion of the first object is within a region-of-interest and a second portion of the first object is outside the region-of-interest in the second overhead depth image. The device is further configured to capture a third overhead depth image of a second object placed on the platform using the 3D sensor at a third time instance. The device is further configured to capture a first image of the second object using a camera in response to determining that the first object is outside of the region-of-interest and the second object is within the region-of-interest for the platform.

Abstract: A device configured to identify a first pixel location within a first plurality of pixels corresponding with an item in a first image and to apply a first homography to the first pixel location to determine a first (x,y) coordinate. The device is further configured to identify a second pixel location within a second plurality of pixels corresponding with the item in a second image and to apply a second homography to the second pixel location to determine a second (x,y) coordinate. The device is further configured to determine that the distance between the first (x,y) coordinate and the second (x,y) coordinate is less than or equal to the distance threshold value, to associate the first plurality of pixels and the second plurality of pixels with a cluster for the item, and to output the first plurality of pixels and the second plurality of pixels.

Abstract: A device configured to receive a first point cloud data for a first item, to identify a first plurality of data points for the first object within the first point cloud data, and to extract the first plurality of data points from the first point cloud data. The device is further configured to receive a second point cloud data for the first item, to identify a second plurality of data points for the first object within the second point cloud data, and to extract a second plurality of data points from the second point cloud data. The device is further configured to merge the first plurality of data points and the second plurality of data points to generate combined point cloud data and to determine dimensions for the first object based on the combined point cloud data.

Abstract: A system for refining an item identification model detects a triggering event at a platform, where the triggering event corresponds to a user placing the item on a platform. The system captures images of the item. The system extracts a set of features from at least one of the images. The system identifies the item based on the set of features. The system receives an indication that the item is not identified correctly. The system receives an identifier of the item. The system identifies the item based on the identifier of the item. The system feeds the identifier of the item and the images to the item identification model. The system retrains the item identification model to learn to associate the item to the images. The system updates the set of features based on the determined association between the item and the images.

Abstract: A device configured to detect a triggering event corresponding with a user placing a first item on the platform, to capture a first image of the first item on the platform using a camera, and to input the first image into a machine learning model that is configured to output a first encoded vector based on features of the first item that are present in the first image. The device is further configured to identify a second encoded vector in an encoded vector library that most closely matches the first encoded vector and to identify a first item identifier in the encoded vector library that is associated with the second encoded vector. The device is further configured to identify the user, to identify an account that is associated with the user, and to associate the first item identifier with the account of the user.

Abstract: A camera array for a scalable tracking system includes cameras that are communicatively coupled to camera clients. The cameras are arranged in a grid such that no camera is directly adjacent in the same row or column of the grid to another camera that is communicatively coupled to the same camera client. Cameras that are arranged along a diagonal of the grid are communicatively coupled to the same camera client.

Abstract: A tracking system includes a camera subsystem that includes cameras that capture vide of a space. Each camera is coupled with a camera client that determines local coordinates of people in the captured video. The camera clients generate frames that include color frames and depth frames labeled with an identifier number of the camera and their corresponding timestamps. The camera clients generate tracks that include metadata describing historical people detections, tracking identifications, timestamps, and the identifier number of the camera. The camera clients send the frames and tracks to cluster servers that maintain the frames and tracks such that they are retrievable using their corresponding labels. A camera server queries the cluster servers to receive the frames and tracks using their corresponding labels. The camera server determines the physical positions of people in the space based on the determined local coordinates.

Abstract: An image sensor is positioned such that a field-of-view of the image sensor encompasses at least a portion of a rack storing items. The image sensor generates images of the items stored on the rack. Over a period of time, a tracking subsystem tracks a pixel position of the wrist of a person interacting with items stored on the rack. receives image frames of the angled-view images. The tracking subsystem determines whether an item was interacted with by a person and, if so, the identified item is assigned to the person.

Abstract: An image sensor is positioned such that a field-of-view of the image sensor encompasses at least a portion of a rack storing items. The image sensor generates angled-view images of the items stored on the rack. A tracking subsystem determines that a person is within a threshold distance of the rack and receives image frames of the angled-view images. A pixel position of a wrist of the person is determined in at least a subset of the received image frames, thereby determining a set of pixel positions of the wrist. An aggregated wrist position is determined based on the set of pixel positions. If the aggregated wrist position is determined to correspond to a position on a shelf of the rack, a trigger signal is provided indicating a shelf-interaction event has occurred.

Abstract: An object tracking system that includes a sensor and a tracking system. The sensor configured to capture a frame of at least a portion of a rack within a global plane for a space. The tracking system is configured to detect an item was removed from the rack. The tracking system is further configured to receive the frame of the rack, to identify a marker on an item within a predefined zone in the frame, and to identify the item associated with the identified marker. The tracking system is further configured to determine a pixel location for a person, to determine the person is within the predefined zone associated with the, and to add the identified item to a digital cart associated with the person.

Abstract: An object tracking system that includes a sensor, a weight sensor, and a tracking system. The sensor configured to capture a frame of at least a portion of a rack within a global plane for a space. The tracking system is configured to detect a weight decrease on the weight sensor. The tracking system is further configured to receive the frame of the rack, to identify a marker on an item within a predefined zone in the frame, and to identify the item associated with the identified marker. The tracking system is further configured to determine a pixel location for a person, to determine the person is within the predefined zone associated with the, and to add the identified item to a digital cart associated with the person.

Abstract: An object tracking system includes a sensor and a tracking system. The sensor is configured to capture a first frame of a global plane for at least a portion of a marker grid in a space. The tracking system is configured to receive a first coordinate in the global plane for a first corner of a marker grid, to determine a second coordinate in the global plane for the first marker on the marker grid, and to determine a third coordinate in the global plane where the second marker on the marker grid. The tracking system is further configured to determine a first pixel location for the first marker, to determine a second pixel location for the second marker, and to generate a homography based on the second coordinate for the first marker, the third coordinate for the second marker, the first pixel location, and the second pixel location.

Abstract: An object tracking system that includes a sensor and a tracking system. The sensor is configured to capture a frame of at least a portion of a rack within a global plane for a space. The tracking system is further configured to receive the frame, to determine a pixel location for the first person, and to determine a person is within the predefined zone associated with the rack. The tracking system is further configured to identify the plurality of items in a digital cart associated with the person, to identify an item from the digital cart associated with the first person, and to remove the identified item from the digital cart associated with the person.

Abstract: An image sensor is positioned such that a field-of-view of the image sensor encompasses at least a portion of a rack storing items. The image sensor generates angled-view images of the items stored on the rack. A tracking subsystem determines that a person is within a threshold distance of the rack and receives image frames of the angled-view images. A pixel position of a wrist of the person is determined in at least a subset of the received image frames, thereby determining a set of pixel positions of the wrist. An aggregated wrist position is determined based on the set of pixel positions. If the aggregated wrist position is determined to correspond to a position on a shelf of the rack, a trigger signal is provided indicating a shelf-interaction event has occurred.

Abstract: A system includes sensors and a tracking subsystem. The subsystem tracks a first object and one or more other objects in a space. After determining that re-identification of the first object is needed, candidate identifiers are determined for the first object. The candidate identifiers include a subset of the identifiers of all tracked objects in the space. The subset includes possible identifiers of the first object based on a history of movements of the first object and interactions of the first object with the other objects in the space. Based on a top-view image, a first descriptor is determined for the first object. The first descriptor is associated with a characteristic of the first object. Based on results of comparing the first descriptor to a set of predetermined descriptors, an updated identifier is determined for the first object.

Abstract: An image sensor is positioned such that a field-of-view of the image sensor encompasses at least a portion of a rack storing items. The image sensor generates angled-view images of the items stored on the rack. A tracking subsystem receives image frames of the angled-view images. The tracking subsystem detects that a trigger event has occurred. A set of one or more image frames from the image feed are determined that are associated with the detected trigger event. A region-of-interest of the image frame is determined based on the pixel position of the wrist of the person. The region-of-interest includes a subset of the pixels of the image frame. A first item in the determined region-of-interest using an object detection algorithm. The identified first item is assigned to the person.

Abstract: An object tracking system that includes a sensor, a weight sensor, and a tracking system. The sensor is configured to capture a frame of at least a portion of a rack within a global plane for a space. The tracking system is configured to detect a weight increase on the weight sensor and to determine a weight increase amount on the weight sensor. The tracking system is further configured to receive the frame, to determine a pixel location for the first person, and to determine a person is within the predefined zone associated with the rack. The tracking system is further configured to identify the plurality of items in a digital cart associated with the person, to identify an item from the digital cart with an item weight that is closest to the weight increase amount, and to remove the identified item from the digital cart associated with the person.

Abstract: A scalable tracking system processes video of a space to track the positions of people within a space. The tracking system determines local coordinates for the people within frames of the video and then assigns these coordinates to time windows based on when the frames were received. The tracking system then combines or clusters certain local coordinates that have been assigned to the same time window to determine a combined coordinate for a person during that time window.