Abstract: A method and related system operations include obtaining a video stream with an image sensor of a camera device, detecting a plurality of target objects by executing a neural network model based on the video stream with a vision processor unit of the camera device. The method also includes generating a plurality of bounding boxes, determining a plurality of character sequences by, for each respective bounding box of the plurality of bounding boxes, performing a set of optical character recognition (OCR) operations to determine a respective character sequence of the plurality of character sequences. The method also includes updating a plurality of tracklets to indicate the plurality of bounding boxes and storing the plurality of tracklets in association with the plurality of character sequences in a memory of the camera device.

Abstract: Automatic license plate recognition occurs when a light sensor that continually captures video detects motion as a vehicle is driven through a gate. The light sensor detects the vehicle and license plate in the video stream captured by the light sensor. An algorithm associated with the video stream of the light sensor is trained to detect license plates. The light sensor starts executing the recognition algorithm when it detects motion. Recognition of characters in the license plate is based upon an aggregation of several captured video frames in which a license plate is detected.

Abstract: A computer vision processor of a camera generates hyperzooms for persons or vehicles from image frames captured by the camera. The hyperzooms include a first hyperzoom associated with the persons or vehicles. The computer vision processor tracks traffic patterns of the persons or vehicles while obviating network usage by the camera by predicting positions of the persons or vehicles using a Kalman Filter from the first hyperzoom. The persons or vehicles are detected in the second hyperzoom. The positions of the persons or vehicles are updated based on detecting the persons or vehicles in the second hyperzoom. The first hyperzoom is removed from the camera. Tracks of the persons or vehicles are generated based on the updated positions. The second hyperzoom is removed from the camera. Track metadata is generated from the tracks for storing in a key-value database located on a non-transitory computer-readable storage medium of the camera.

Abstract: A method includes building a firmware image to execute on a bootloader of a system on chip (SoC), the firmware image including first encryption public and private keys, and digitally signing the firmware image with a second encryption private key. The signed firmware image is encrypted with a symmetric encryption key, which in turn is encrypted with a second encryption public key. The encrypted signed firmware image and the encrypted symmetric encryption key are sent to the SoC to cause the SoC to (1) decrypt the encrypted symmetric encryption key to produce the symmetric encryption key using a third encryption private key from a first asymmetric key pair, (2) decrypt the encrypted signed firmware image to produce the signed firmware image using the symmetric encryption key, and (3) verify a digital signature of the signed firmware image using a third encryption public key from a second asymmetric key pair.

Abstract: A first set of values reported by an electronic device and not reported by another electronic device over a first period of time that is prior to a firmware release to the electronic device is received. The first set of values is associated with a metric. A set of statistical properties associated with the first set of values is determined. A second set of values reported by the electronic device and not reported by another electronic device over a second period of time that is after the firmware release is received. The second set of values is associated with the metric. A set of statistical properties associated with the second set of values is determined. The set of statistical properties associated with the first set of values and the set of statistical properties associated with the second set of values is compared to detect an anomaly.

Abstract: A computer vision processor of a camera extracts attributes of persons or vehicles from hyperzooms generated from image frames. The hyperzooms represent traffic patterns. The extracting is performed using a feature extractor of an on-camera convolutional neural network (CNN) including an inverted residual structure. The attributes include at least colors of clothing of the persons or colors of the vehicles. Mobile semantic segmentation models of the CNN are generated using the hyperzooms and the attributes. Attribute analytics are generated by executing the mobile semantic segmentation models while obviating network usage by the camera. The attribute analytics are stored in a key-value database located on a memory card of the camera. A query is received from the server instance specifying one or more of the attributes. The attribute analytics are filtered using the one or more of the attributes to obtain a portion of the traffic patterns.

Abstract: A method includes generating a first representative vector based on a first vectors, wherein the first representative vector is associated with the first vectors in a collection of representative vectors, and the first vectors comprises a set of vector values within a latent space. The method further includes generating a second representative vector based on a second vectors, wherein the second representative vector is associated with the second vectors in the collection of representative vectors. The method further includes determining a latent space distance based on the first and second vectors. The method further includes determining whether the latent space distance satisfies a threshold. In response to a determination that the latent space distance satisfies the threshold, the method further includes associating a combined representative vector with the first vectors and the second vectors and removing the first and second representative vectors from the collection of representative vectors.

Abstract: A computer vision processor of a camera generates hyperzooms for persons or vehicles from image frames captured by the camera. The hyperzooms include a first hyperzoom associated with the persons or vehicles. The computer vision processor tracks traffic patterns of the persons or vehicles while obviating network usage by the camera by predicting positions of the persons or vehicles using a Kalman Filter from the first hyperzoom. The persons or vehicles are detected in the second hyperzoom. The positions of the persons or vehicles are updated based on detecting the persons or vehicles in the second hyperzoom. The first hyperzoom is removed from the camera. Tracks of the persons or vehicles are generated based on the updated positions. The second hyperzoom is removed from the camera. Track metadata is generated from the tracks for storing in a key-value database located on a non-transitory computer-readable storage medium of the camera.

Abstract: A computer vision processor of a camera extracts attributes of persons or vehicles from hyperzooms generated from image frames. The hyperzooms represent traffic patterns. The extracting is performed using a feature extractor of an on-camera convolutional neural network (CNN) including an inverted residual structure. The attributes include at least colors of clothing of the persons or colors of the vehicles. Mobile semantic segmentation models of the CNN are generated using the hyperzooms and the attributes. Attribute analytics are generated by executing the mobile semantic segmentation models while obviating network usage by the camera. The attribute analytics are stored in a key-value database located on a memory card of the camera. A query is received from the server instance specifying one or more of the attributes. The attribute analytics are filtered using the one or more of the attributes to obtain a portion of the traffic patterns.

Abstract: Automatic license plate recognition occurs when a light sensor that continually captures video detects motion as a vehicle is driven through a gate. The light sensor detects the vehicle and license plate in the video stream captured by the light sensor. An algorithm associated with the video stream of the light sensor is trained to detect license plates. The light sensor starts executing the recognition algorithm when it detects motion. Recognition of characters in the license plate is based upon an aggregation of several captured video frames in which a license plate is detected.

Abstract: A method includes generating a first representative vector based on a first vectors, wherein the first representative vector is associated with the first vectors in a collection of representative vectors, and the first vectors comprises a set of vector values within a latent space. The method further includes generating a second representative vector based on a second vectors, wherein the second representative vector is associated with the second vectors in the collection of representative vectors. The method further includes determining a latent space distance based on the first and second vectors. The method further includes determining whether the latent space distance satisfies a threshold. In response to a determination that the latent space distance satisfies the threshold, the method further includes associating a combined representative vector with the first vectors and the second vectors and removing the first and second representative vectors from the collection of representative vectors.

Abstract: A method includes obtaining a plurality of representative vectors associated with face-related data. The method includes determining an encryption key based on a parameter stored in a record, generating an encrypted vector set by, for each respective vector of the plurality of representative vectors, encrypting the respective vector with a homomorphic encryption operation based on the encryption key, where the encrypted vector set includes a first encrypted vector that is linked to a subset of the face-related data associated with the first plurality of face vectors. The method further includes obtaining an encrypted face search vector using the encryption key to perform homomorphic encryption. The method further includes selecting a first encrypted vector based on the encrypted face search vector and retrieving the subset of the face-related data based on the first encrypted vector.

Abstract: A computer vision processor of a camera extracts attributes of persons or vehicles from hyperzooms generated from image frames. The hyperzooms represent traffic patterns. The extracting is performed using a feature extractor of an on-camera convolutional neural network (CNN) including an inverted residual structure. The attributes include at least colors of clothing of the persons or colors of the vehicles. Mobile semantic segmentation models of the CNN are generated using the hyperzooms and the attributes. Attribute analytics are generated by executing the mobile semantic segmentation models while obviating network usage by the camera. The attribute analytics are stored in a key-value database located on a memory card of the camera. A query is received from the server instance specifying one or more of the attributes. The attribute analytics are filtered using the one or more of the attributes to obtain a portion of the traffic patterns.