Abstract: An action recognition method includes: obtaining original feature submaps of each of temporal frames on a plurality of convolutional channels by using a multi-channel convolutional layer; calculating, by using each of the temporal frames as a target temporal frame, motion information weights of the target temporal frame on the convolutional channels according to original feature submaps of the target temporal frame and original feature submaps of a next temporal frame, and obtaining motion information feature maps of the target temporal frame on the convolutional channels according to the motion information weights; performing temporal convolution on the motion information feature maps of the target temporal frame to obtain temporal motion feature maps of the target temporal frame; and recognizing an action type of a moving object in image data of the target temporal frame according to the temporal motion feature maps of the target temporal frame on the convolutional channels.

Abstract: Disclosed herein are an image detection method and apparatus, a computer-readable storage medium, and a computer device. The method includes iteratively training a plurality of neural network models to obtain a plurality of trained neural network model; and performing detection on an image to be detected using the trained plurality of neural network models to obtain a detection result. Each iteration of training includes: for each of a plurality of sample images, separately inputting the sample image into the neural network models to obtain a fuzzy probability value set, and calculating, based on the fuzzy probability value set and preset label information of the sample image, a loss parameter of the sample image; selecting target sample images based on a distribution of loss parameters of the plurality of sample images; and updating the plurality of neural network models based on the target sample images.

Abstract: An image processing method and apparatus that includes: performing feature extraction processing on N image regions of a to-be-processed image respectively to obtain feature data corresponding to the N image regions respectively, N being an integer greater than or equal to 1, performing defect detection on the N image regions respectively according to the feature data corresponding to the N image regions respectively to obtain a prediction probability of a defect in each of the N image regions, obtaining attention for each image region, adjusting the prediction probability of the defect in each image region according to the attention for each image region, and generating a prediction result of the to-be-processed image according to the adjusted prediction probability of each image region.

Abstract: In the field of artificial intelligence technologies, a gaze correction method and apparatus for a face image, a device, a computer-readable storage medium, and a computer program product are provided. The method includes: acquiring an eye image from a face image; determining an eye movement flow field based on the eye image and a target gaze direction, the target gaze direction being a gaze direction to which an eye gaze in the eye image is to be corrected; adjusting a pixel position in the eye image based on the eye movement flow field, to obtain a corrected eye image; and generating a face image with a corrected gaze based on the corrected eye image.

Abstract: The present subject matter discloses an action recognition method, apparatus and device, a storage medium, and a computer program product, belonging to the field of image recognition. Multiple video frames in a target video are obtained. Feature extraction is performed on the multiple video frames respectively according to multiple dimensions to obtain multiple multi-channel feature patterns. Each video frame corresponds to one multi-channel feature pattern. Each channel represents one dimension. An attention weight of each multi-channel feature pattern is determined based on a similarity between every two multi-channel feature patterns. The attention weight is used for representing a degree of correlation between a corresponding multi-channel feature pattern and an action performed by an object in the target video. A type of the action is determined based on the multiple multi-channel feature patterns and the determined multiple attention weights.

Abstract: An image gaze correction method, apparatus, electronic device, computer-readable storage medium, and computer program product. The image gaze correction method includes: acquiring a to-be-corrected eye image from a to-be-corrected image, generating, based on the to-be-corrected eye image, an eye motion flow field and an eye contour mask, the eye motion flow field being used for adjusting a pixel position in the to-be-corrected eye image, and the eye contour mask being used for indicating a probability that the pixel position in the to-be-corrected eye image belongs to an eye region, performing, based on the eye motion flow field and the eye contour mask, gaze correction processing on the to-be-corrected eye image to obtain a corrected eye image, and generating a gaze corrected image based on the corrected eye image.

Abstract: An image gaze correction method, apparatus, electronic device, computer-readable storage medium, and computer program product related to the field of artificial intelligence technologies are provided. The image gaze correction method includes: acquiring an eye image from an image; performing feature extraction processing on the eye image to obtain feature information of the eye image; performing, based on the feature information and a target gaze direction, gaze correction processing on the eye image to obtain an initially corrected eye image and an eye contour mask; performing, by using the eye contour mask, adjustment processing on the initially corrected eye image to obtain a corrected eye image; and generating a gaze corrected image based on the corrected eye image.

Abstract: This present disclosure describes a video image processing method and apparatus, a computer-readable medium and an electronic device, relating to the field of image processing technologies. The method includes determining, by a device, a target-object region in a current frame in a video. The device includes a memory storing instructions and a processor in communication with the memory. The method also includes determining, by the device, a target-object tracking image in a next frame and corresponding to the target-object region; and sequentially performing, by the device, a plurality of sets of convolution processing on the target-object tracking image to determine a target-object region in the next frame. A quantity of convolutions of a first set of convolution processing in the plurality of sets of convolution processing is less than a quantity of convolutions of any other set of convolution processing.

Abstract: A face detection method includes: acquiring a target image; invoking a face detection network, and processing the target image by using a feature extraction structure of the face detection network, to obtain original feature maps corresponding to the target image; the original feature maps having different resolutions; processing the original feature maps by using a feature enhancement structure of the face detection network, to obtain an enhanced feature map corresponding to each original feature map; the feature enhancement structure being obtained by searching a search space, and the search space used for searching the feature enhancement structure being determined based on a detection objective of the face detection network and a processing object of the feature enhancement structure; and processing the enhanced feature map by using a detection structure of the face detection network, to obtain a face detection result of the target image.

Abstract: An action recognition method includes: obtaining original feature submaps of each of temporal frames on a plurality of convolutional channels by using a multi-channel convolutional layer; calculating, by using each of the temporal frames as a target temporal frame, motion information weights of the target temporal frame on the convolutional channels according to original feature submaps of the target temporal frame and original feature submaps of a next temporal frame, and obtaining motion information feature maps of the target temporal frame on the convolutional channels according to the motion information weights; performing temporal convolution on the motion information feature maps of the target temporal frame to obtain temporal motion feature maps of the target temporal frame; and recognizing an action type of a moving object in image data of the target temporal frame according to the temporal motion feature maps of the target temporal frame on the convolutional channels.

Abstract: A trajectory tracking method is provided for a computer device. The method includes performing motion tracking on head images in a plurality of video frames, to obtain motion trajectories corresponding to the head images; acquiring face images corresponding to the head images in the video frames, to obtain face image sets corresponding to the head images; determining from the face image sets corresponding to the head images, at least two face image sets having same face images; and combining motion trajectories corresponding to the at least two face image sets having same face images, to obtain a final motion trajectory of trajectory tracking.

Abstract: When a target image is captured, the device provides a portion of the target image within a target detection region to a preset first model set to calculate positions of face key points and a first confidence value. The face key points and the first confidence value are output by the first model set for a single input of the portion of the first target image into the first model set. When the first confidence value meets a first threshold corresponding to whether the target image is a face image, the device obtains a second target image corresponding to the positions of the first face key points; the device inputs the second target image into the first model set to calculate a second confidence value, the second confidence value corresponds to accuracy key point positioning, and outputs the first key points if the second confidence value meets a second threshold.

Abstract: A facial tracking method is provided. The method includes: obtaining, from a video stream, an image that currently needs to be processed as a current image frame; and obtaining coordinates of facial key points in a previous image frame and a confidence level corresponding to the previous image frame. The method also includes calculating coordinates of facial key points in the current image frame according to the coordinates of the facial key points in the previous image frame when the confidence level is higher than a preset threshold; and performing multi-face recognition on the current image frame according to the coordinates of the facial key points in the current image frame. The method also includes calculating a confidence level of the coordinates of the facial key points in the current image frame, and returning to process a next frame until recognition on all image frames is completed.

Abstract: This present disclosure describes a video image processing method and apparatus, a computer-readable medium and an electronic device, relating to the field of image processing technologies. The method includes determining, by a device, a target-object region in a current frame in a video. The device includes a memory storing instructions and a processor in communication with the memory. The method also includes determining, by the device, a target-object tracking image in a next frame and corresponding to the target-object region; and sequentially performing, by the device, a plurality of sets of convolution processing on the target-object tracking image to determine a target-object region in the next frame. A quantity of convolutions of a first set of convolution processing in the plurality of sets of convolution processing is less than a quantity of convolutions of any other set of convolution processing.

Abstract: A trajectory tracking method is provided for a computer device. The method includes performing motion tracking on head images in a plurality of video frames, to obtain motion trajectories corresponding to the head images; acquiring face images corresponding to the head images in the video frames, to obtain face image sets corresponding to the head images; determining from the face image sets corresponding to the head images, at least two face image sets having same face images; and combining motion trajectories corresponding to the at least two face image sets having same face images, to obtain a final motion trajectory of trajectory tracking.

Abstract: A facial tracking method is provided. The method includes: obtaining, from a video stream, an image that currently needs to be processed as a current image frame; and obtaining coordinates of facial key points in a previous image frame and a confidence level corresponding to the previous image frame. The method also includes calculating coordinates of facial key points in the current image frame according to the coordinates of the facial key points in the previous image frame when the confidence level is higher than a preset threshold; and performing multi-face recognition on the current image frame according to the coordinates of the facial key points in the current image frame. The method also includes calculating a confidence level of the coordinates of the facial key points in the current image frame, and returning to process a next frame until recognition on all image frames is completed.

Abstract: When a target image is captured, the device provides a portion of the target image within a target detection region to a preset first model set to calculate positions of face key points and a first confidence value. The face key points and the first confidence value are output by the first model set for a single input of the portion of the first target image into the first model set. When the first confidence value meets a first threshold corresponding to whether the target image is a face image, the device obtains a second target image corresponding to the positions of the first face key points; the device inputs the second target image into the first model set to calculate a second confidence value, the second confidence value corresponds to accuracy key point positioning, and outputs the first key points if the second confidence value meets a second threshold.