A control method for a mobile platform includes obtaining a captured image, determining a target first characteristic part of a target object from the captured image, determining a second characteristic part of the target object in the captured image, and switching from tracking the second characteristic part to tracking the target first characteristic part in response to a tracking parameter of the target object meeting a preset tracking condition.

The learning device includes a first trainer, a candidate area determinator and a second trainer. On the basis of a training image and a training label including the correct coordinate value relating to the feature point included in the training image, the first trainer generates a first discriminator learned to output the predicted coordinate value relating to a feature point from an input image. The candidate area determinator determines a candidate area of the feature point in the training image based on the predicted coordinate value outputted by inputting the training image to the first discriminator. The second trainer generates the second discriminator, which is learned to output the reliability map indicating the reliability to the feature point at each block in an input image, on the basis of a cut-out image that is the candidate area cut out from the training image.

The present disclosure provides an image processing method. An image to be classified is input into a feature extraction model to generate N dimensional features. Dimension fusion is performed on M features of the N dimensional features to obtain M dimension fusion features. The image to be classified is processed based on M dimension fusion features and remaining features of the N dimensional features other than the M features.

A video processing method and apparatus is provided, the video processing method includes: dividing a received initial video into at least one video segment; obtaining, based on a feature extraction model, a first modal feature, a second modal feature, and a third modal feature that correspond to each video segment in the at least one video segment; and inputting, into a recognition model, the first modal feature, the second modal feature, and the third modal feature that correspond to each video segment, to obtain a video score corresponding to each video segment, and determining a target video segment in the initial video based on the video score.

A video processing method and apparatus are provided. The video processing method includes: extracting at least two types of modal information from a received target video; extracting, based on a preset feature extraction model, at least two modal features corresponding to the at least two types of modal information; and fusing the at least two modal features to obtain a target feature of the target video.

Provided is a deep learning-based panoptic segmentation accelerated processing technique using a complexity-based RPN skip method. An image segmentation system includes: a first processing unit configured to extract dynamic objects in an instance segmentation method by using an extracted feature; a calculation unit configured to control to skip some areas of the feature extracted at the network by the first processing unit, on the basis of complexity of the input image; a second processing unit configured to extract static objects in a semantic segmentation method by using the feature extracted at the network; and a fusion unit configured to fuse a result of extracting by the first processing unit and a result of extracting by the second processing unit. Accordingly, the panoptic segmentation method can be easily performed even in an embedded environment by reducing complexity for panoptic segmentation processing by reducing a calculation burden.

Systems and techniques are described for image processing. An imaging system receives an identity image and an attribute image. The identity image depicts a first person having an identity. The attribute image depicts a second person having an attribute, such as a facial feature, an accessory worn by the second person, and/or an expression. The imaging system uses trained machine learning model(s) to generate a combined image based on the identity image and the attribute image. The combined image depicts a virtual person having both the identity of the first person and the attribute of the second person. The imaging system outputs the combined image, for instance by displaying the combined image or sending the combined image to a receiving device. In some examples, the imaging system updates the trained machine learning model(s) based on the combined image.

The described aspects and implementations enable efficient calibration of a sensing system of a vehicle. In one implementation, disclosed is a method and a system to perform the method, the system including the sensing system configured to collect sensing data, characterizing an environment of the vehicle, the sensing data including infrared sensing data. The system further includes a data processing system operatively coupled to the sensing system and configured to process the sensing data using a classifier machine-learning model to obtain a classification of one or more vulnerable road users present in the environment of the vehicle.

One embodiment can provide a computer-vision system. The computer-vision system can include one or more cameras to capture images of a scene and one or more sets of single-color light sources to illuminate the scene, with a respective set of light sources comprising multiple single-color light sources of different colors. The multiple single-color light sources within a given set can be turned on sequentially, one at a time. The cameras can capture an image of the scene each time the scene is illuminated by a respective single-color light source of a particular color.

Methods, apparatuses, electronic devices, and computer readable storage media for image processing are provided. In one aspect, an image processing method includes: determining a plurality of image feature maps of a target image, the plurality of image feature maps corresponding to different preset scales; determining, based on the plurality of image feature maps and for each pixel of pixels in the target image, a first probability that the pixel in the target image belongs to a foreground and a second probability that the pixel in the target image belongs to a background; and performing panoramic segmentation on the target image based on the plurality of image feature maps, the first probabilities of the pixels in the target image, and the second probabilities of the pixels in the target image.