A robotic system includes one or more optical sensors configured to separately obtain two dimensional (2D) image data and three dimensional (3D) image data of a brake lever of a vehicle, a manipulator arm configured to grasp the brake lever of the vehicle, and a controller configured to compare the 2D image data with the 3D image data to identify one or more of a location or a pose of the brake lever of the vehicle. The controller is configured to control the manipulator arm to move toward, grasp, and actuate the brake lever of the vehicle based on the one or more of the location or the pose of the brake lever.

In some embodiments, systems and methods are provided to recognize retail products, comprising: a model training system configured to: identify a customer; access an associated customer profile; access and apply a set of filtering rules to a product database based on customer data; generate a listing of products specific to the customer; access and apply a model training set of rules to train a machine learning model based on the listing of products and corresponding image data for each of the products in the listing of products; and communicate the trained model to the portable user device associated with first customer.

Certain implementations of the subject matter can be implemented as a method of screening demulsifiers for live crude oil-water emulsions. A live emulsion of a hydrocarbon sample and a water sample is flowed through a capillary viscometer. The live emulsion includes dissolved gases retrieved from a hydrocarbon-carrying reservoir. While flowing the live emulsion through the capillary viscometer, a demulsifier sample is flowed through the capillary viscometer. The demulsifier sample causes breakdown of the live emulsion. Using the capillary viscometer, change in a viscosity of the live emulsion over time resulting from the breakdown of the live emulsion due to the demulsifier sample is measured. Multiple images of the breakdown of the live emulsion over time are captured. A strength of the live emulsion is classified based, in part, on the change in the viscosity of the live emulsion over time and on the plurality of images.

An image sensor for reducing channel variation and an image processing system including the same. The image sensor includes first to m.sup.th pixels (m≧2), each of which is connected to a corresponding column line from among first to m.sup.th column lines and is configured to output a respective pixel signal.’ The image sensor further includes first to m.sup.th bias circuits, each of which is connected to a corresponding column line from among the first to m.sup.th column lines and is configured to fix a voltage of the corresponding column line to a bias voltage when a column line-specific pixel is not selected to output the respective pixel signal. An analog-to-digital converter in the image sensor is configured to convert the pixel signals into digital signals.

A method includes receiving first tracking data and a first confidence value at a tracker selection system from a first tracking system. The method includes receiving second tracking data and a second confidence value at the tracker selection system from a second tracking system. The tracking systems may be configured to track an object in a sequence of images, and the first and second tracking data may indicate locations of regions where the corresponding tracking system has tracked the object in an image of the sequence of images. The confidence values may indicate likelihoods that a corresponding tracking system is tracking the object. The method further includes providing output data to the first tracking system and to the second tracking system. The output data may include data selected from the first tracking data and the second tracking data based on a result of a comparison of the confidence values.

A method and an apparatus for classifying a video are provided. The method may include: acquiring a to-be-classified video; extracting a set of multimodal features of the to-be-classified video; inputting the set of multimodal features into a post-fusion model corresponding to each modal respectively, to obtain multimodal category information of the to-be-classified video; and fusing the multimodal category information of the to-be-classified video, to obtain category information of the to-be-classified video. This embodiment improves the accuracy of video classification.

Systems and methods are described for Distification of 3D imagery. A computing device may obtain a three dimensional (3D) image that defines a 3D point cloud used to generate a two dimensional (2D) image matrix. The 2D image matrix may include 2D matrix point(s), where each 2D matrix point can be associated with a horizontal coordinate and a vertical coordinate. The computing device can generate an output feature vector that includes at least one 2D matrix point of the 2D image matrix, and a 3D point in the 3D point cloud of the 3D image. The 3D point in the 3D point cloud is mapped to a coordinate pair comprised of the horizontal coordinate and the vertical coordinate of the at least one 2D matrix point of the 2D image matrix point. The output feature vector is input into a predictive model.

A system and method for monitoring vehicle traffic and collecting data indicative of pedestrian right of way violations by vehicles is provided. The system comprises memory and logic for monitoring traffic intersections and recording evidence indicating that vehicles have violated pedestrian right of way. Two sensor modalities collecting video data and radar data of the intersection under observation are employed in one embodiment of the system. The violation evidence can be accessed remotely by a traffic official for issuing of traffic citations.

An improved system and method of determining presence of stop signs at intersection based on multiple types of data including ground truth data, telemetry data, and/or trip stills gathered from vehicle devices. The data may be analyzed and aggregated by one or more models and/or neural networks to develop a prediction regarding presence of stop signs at individual intersections.

A multimodal and real-time method for filtering sensitive content, receiving as input a digital video stream, the method including segmenting digital video into video fragments along the video timeline; extracting features containing significant information from the digital video input on sensitive media; reducing the semantic difference between each of the low-level video features, and the high-level sensitive concept; classifying the video fragments, generating a high-level label (positive or negative), with a confidence score for each fragment representation; performing high-level fusion to properly match the possible high-level labels and confidence scores for each fragment; and predicting the sensitive time by combining the labels of the fragments along the video timeline, indicating the moments when the content becomes sensitive.