Systems, methods, and instrumentalities are disclosed for computer vision-based surgical workflow recognition using natural language processing (NLP) techniques. Surgical video of surgical procedures may be processed and analyzed, for example, to achieve workflow recognition. Surgical phases may be determined based on the surgical video and segmented to generate an annotated video representation. The annotated video representation of the surgical video may provide information associated with the surgical procedure. For example, the annotated video representation may provide information on surgical phases, surgical events, surgical tool usage, and/or the like.

When three-dimensional audio is produced by using headphones, particular HRTF-filters are used to modify sound for the left and right channels of the headphone. As the morphology of every ear is different, it is beneficial to have HRTF-filters particularly designed for the user of headphones. Such filters may be produced by deriving ear geometry from a plurality of images taken with an ordinary camera, detecting necessary features from images and fitting said features to a model that has been produced from accurately scanned ears comprising representative values for different sizes and shapes. Taken images are sent to a server (52) that performs the necessary computations and submits the data further or produces the requested filter.

An image information overlay system retrieves an image associated with a vehicle listing and uses machine learning models to classify the image, generating identification data that may comprise a vehicle make and model, a feature or part of the vehicle present in the image, and a location of the vehicle feature or part. The identification data or an individual identifier of the vehicle, such as a Vehicle Identification Number (VIN), may be used to retrieve overlay information related to the vehicle make and model, such as recalls or known maintenance issues or information specific to the vehicle, such as mileage, accident reports, or ownership history. The overlay information is displayed on the image as an overlay at the location of the vehicle feature or part corresponding to the overlay information.

An inspection apparatus according to one or more embodiments extracts an attention area from a target image using a first estimation model, performs a computational process with a second estimation model using the extracted attention area, and determines whether a target product has a defect based on a computational result from the second estimation model. The first estimation model is generated based on multiple first training images of defect-free products in a target environment. The second estimation model is generated based on multiple second training images of defects. The computational process with the second estimation model includes generating multiple feature maps with different dimensions by projecting the target image into different spaces with lower dimensions. The extracted attention area is integrated into at least one of the multiple feature maps in the computational process with the second estimation model.

A method, apparatus and system for 3D object detection and segmentation are provided. The method comprises the steps of: extracting multi-view 2D features based on multi-view images captured by a plurality of cameras; generating a 3D feature volume based on the multi-view 2D features; and performing a depth estimation, a semantic segmentation, and a 3D object detection based on the 3D feature volume. The method, apparatus, and system of the disclosure are faster, computation friendly, flexible, and more practical to deploy on vehicles, drones, robots, vehicles, mobile devices, or mobile communication devices.

A counterfeit and imaging detection system includes a processor, a counterfeit product detection app, and a steganographic imaging model, electronically accessible by the counterfeit product detection app trained using image data and configured cause the processor to obtain a digital image of a physical product of a product line, the digital image captured by an imaging device and the digital image comprising pixel data, analyze the digital image to detect within the pixel data a batch code uniquely identifying a batch of the physical product of the product line, analyze the pixel data of the digital image to determine that the batch code is counterfeit, and augment a counterfeit list of batch codes to include the batch code, wherein the counterfeit list of batch codes remains electronically accessible to the counterfeit product detection app for one or more further counterfeit detection iterations.

The present disclosure discloses a multiscale point cloud classification method. The method includes the following steps: acquiring 3D unordered point cloud data; performing feature extraction and classification on the acquired point cloud data using a pre-trained parallel classification network to obtain an output result, wherein the parallel classification network includes a plurality of basic networks with the same structures; and fusing the output results of the parallel network using a pre-trained deep Q network to obtain a final result of point cloud classification. The present disclosure can improve the accuracy and robustness of point cloud classification.

A method and apparatus for object recognition are provided. A processor-implemented method includes extracting feature maps including local feature representations from an input image, generating a global feature representation corresponding to the input image by fusing the local feature representations, and performing a recognition task on the input image based on the local feature representations and the global feature representation.

The present invention proposes an ensemble deep learning method for identifying unsafe behaviors of operators in maritime working environment. Firstly, extract features of maritime images with the You Only Look Once (YOLO) V3 model, and then enhance a multi-scale detection capability by introducing a feature pyramid structure. Secondly, obtain instance-level features and time memory features of the operators and devices in the maritime working environment with the Joint Learning of Detection and Embedding (JDE) paradigm. Thirdly, transfer spatial-temporal interaction information into a feature memory pool, and update the time memory features with the asynchronous memory updating algorithm. Finally, identify the interaction between the operators, the devices, and unsafe behaviors with an asynchronous interaction aggregation network. The proposed invention can accurately determine the unsafe behaviors of the operators, and thus provide operation decisions for maritime management relevant activities.

Systems and methods presented herein are configured to train a neural network model using a first set of photographs, wherein each photograph of the first set of photographs depicts a first set of objects and include one or more annotations relating to each object of the first set of objects; to automatically create mask images corresponding to a second set of objects depicted by a second set of photographs; to enable manual fine tuning of the mask images corresponding to the second set of objects depicted by the second set of photographs; to re-train the neural network model using the second set of photographs, wherein the re-training is based at least in part on the manual fine tuning of the mask images corresponding to the second set of objects depicted by the second set of photographs; and to identify one or more individual objects in a third set of photographs using the re-trained neural network model.