Abstract: Embodiments of the disclosure relate method, apparatus, device and storage medium for processing a media content. The method includes: obtaining a first media content; and generating a second media content by applying an effect to the first media content with a model, wherein the model is trained based on the following process: determining a plurality of sub-effects corresponding to the effect; determining a plurality of effect models corresponding to the plurality of sub-effects from a set of pre-trained effect model; generating a plurality of corresponding output images by processing a plurality of sample images with the plurality of effect models according to a preset order; and training the model with the plurality of sample images and the plurality of corresponding output images. According to embodiments of the present disclosure, the generation efficiency of the model can be improved.

Abstract: Embodiments of the disclosure relate to a method, a device, an electronic device and a storage medium for processing media content. The method includes: obtaining first media content; applying a first effect to the first media content by a first model to generate second media content; and providing the second media content, wherein the first model is trained by: training a pre-trained model associated with the first effect with a first sample set to determine a reference model; constructing a second sample set based on processing results for a plurality of sample images from the reference model; and training the first model with the second sample set.

Abstract: The embodiment of the disclosure relates to a method, apparatus, device and storage medium for generating decoration element. The method includes: obtaining input information indicating an element category and an element style; providing a target decoration element generated based on the input information, wherein the target decoration element is generated by: determining mask information corresponding to the element category, the mask information indicating a target area for generating a decoration element; providing a reference image, the mask information, and a first prompt corresponding to the input information to a first target model to indicate the first target model to fill the target area in the reference image; and generating the target decoration element based on a filling result of the target area. According to the embodiment of the disclosure, the quality of the generated decoration element can be improved.

Abstract: The present disclosure relates to an effect processing method, an electronic device and a non-transitory storage medium. The effect processing method includes: in response to an effect processing request input on a mobile terminal, acquiring an image to be processed; and performing effect processing on the image to be processed by a first generation model deployed on the mobile terminal to obtain a target effect image, and displaying the target effect image; the first generation model is obtained by training a first generative adversarial network, at least part of training data of the first generative adversarial network is generated by a second generation model deployed on a server, and the second generation model is obtained by training a second generative adversarial network.

Abstract: The embodiment of the disclosure relates to a method, apparatus, device and a computer readable storage medium of processing information. The method proposed herein includes: obtaining target content to be processed; determining a target encoding representation of the target content with an encoding model; and determining a target generation manner of the target content based on a comparison between the target encoding representation and a plurality of predetermined encoding representations, the plurality of predetermined encoding representations corresponding to a plurality of predetermined generation manners, the plurality of predetermined generation manners including a plurality of model generation manners, the plurality of predetermined encoding representations being determined by processing a plurality of groups of sample contents with the encoding model, each group of sample contents corresponding to a respective predetermined generation manner.

Abstract: A dolly zoom effect can be applied to one or more images captured via a resource-constrained device (e.g., a mobile smartphone) by manipulating the size of a target feature while the background in the one or more images changes due to physical movement of the resource-constrained device. The target feature can be detected using facial recognition or shape detection techniques. The target feature can be resized before the size is manipulated as the background changes (e.g., changes perspective).

Abstract: A dolly zoom effect can be applied to one or more images captured via a resource-constrained device (e.g., a mobile smartphone) by manipulating the size of a target feature while the background in the one or more images changes due to physical movement of the resource-constrained device. The target feature can be detected using facial recognition or shape detection techniques. The target feature can be resized before the size is manipulated as the background changes (e.g., changes perspective).

Abstract: Disclosed are systems, methods, and computer-readable storage media to enhance an image on a mobile computing device. In some aspects, a user interface on the mobile computing device provides a plurality of image filters, each of the image filters associated with a different replacement sky type. Upon application of one of the plurality of image filters to an image, the mobile computing device is configured to identify a sky portion of the image, modify a color mapping of a non-sky portion of the image, and replace the sky portion of the image with replacement sky data corresponding to the applied image filter.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.

Abstract: Disclosed are systems, methods, and computer-readable storage media to modify image content. One aspect includes identifying, by one or more electronic hardware processors, an image and content within the image, determining, by the one or more electronic hardware processors, a sky region of the image, determining, by the one or more electronic hardware processors, whether the content within the image is located within the sky region of the image, and in response to the content being within the sky region of the image, modifying, by the one or more electronic hardware processors, the content based on fractal Brownian motion.

Abstract: A neural network training method, an image processing method, and apparatuses thereof are provided. The neural network training method includes obtaining a first domain image and a second domain image, where the first domain image and the second domain image are unpaired images in different domains; obtaining a scaled first domain image by scaling, at an iteration, the first domain image; obtaining a training patch by cropping the scaled first domain image, where each training patch has a same number of pixels with different contents; inputting the training patch into the neural network at the iteration, and outputting an output patch; calculating a contrastive loss based on a query sub-patch and negative sub-patches selected from the training patch and a corresponding positive sub-patch selected from the output patch; and updating model parameters of the neural network based on the contrastive loss and a generative adversarial network loss.

Abstract: A method for processing an image, an electronic device and a storage medium are provided. The method includes: obtaining an original image including a target object; obtaining an auxiliary line by extracting semantic information from the original image, the auxiliary line including at least one of: an area boundary line of the target object and a part contour line of the target object; obtaining a prediction result for a semantic line by inputting an image stitched from the auxiliary line and the original image to a predictive neural network, the auxiliary line guiding the predictive neural network to obtain the prediction result, the prediction result indicating a probability that a pixel in the original image is a pixel in the semantic line, the semantic line being used for rendering the target object; and obtaining the semantic line based on the prediction result for the semantic line.

Abstract: A hair rendering method includes acquiring a target video containing hair information, and selecting a target image frame from image frames of the target video, acquiring a texture image of the target image frame, wherein the texture image is an image in a texture format which records motion states and state change information of one or more pixel points in the target image frame. The method further includes acquiring a first target motion state and target state change information of a first target particle region from the texture image; determining a second target motion state of the first target particle region in a next image frame according to the first target motion state and the target state change information, and rendering the hair region by updating a motion state in the texture image according to the second target motion state.

Abstract: A dolly zoom effect can be applied to one or more images captured via a resource-constrained device (e.g., a mobile smartphone) by manipulating the size of a target feature while the background in the one or more images changes due to physical movement of the resource-constrained device. The target feature can be detected using facial recognition or shape detection techniques. The target feature can be resized before the size is manipulated as the background changes (e.g., changes perspective).

Abstract: Various embodiments provide for systems, methods, and computer-readable storage media for annotating a digital image with a texture fill. An annotation system may receive a user input defining a border separating a first portion of a target digital image from a second portion of the target digital image. The annotation system may then generate a contour mask, such as a binary mask, for the target digital image based on the user-defined border. The annotation system may then apply a media overlay to the target image based on the contour mask. In particular, the contour mask can define portions of the target digital image such that the annotation system will apply a media overlay to at least one of those portions while not applying the media overlay to remaining portions of the target digital image.

Abstract: A dolly zoom effect can be applied to one or more images captured via a resource-constrained device (e.g., a mobile smartphone) by manipulating the size of a target feature while the background in the one or more images changes due to physical movement of the resource-constrained device. The target feature can be detected using facial recognition or shape detection techniques. The target feature can be resized before the size is manipulated as the background changes (e.g., changes perspective).

Abstract: A method for processing images includes recognizing a hair region in a target image; acquiring a hairline direction of each pixel point in the hair region; acquiring an intermediate image by smoothing the hair region based on the hairline direction; and acquiring a virtual hairline along the hairline direction, and acquiring a first image by superimposing the virtual hairline to the hair region in the intermediate image.

Abstract: Disclosed are systems, methods, and computer-readable storage media to enhance an image on a mobile computing device. In some aspects, a user interface on the mobile computing device provides a plurality of image filters, each of the image filters associated with a different replacement sky type. Upon application of one of the plurality of image filters to an image, the mobile computing device is configured to identify a sky portion of the image, modify a color mapping of a non-sky portion of the image, and replace the sky portion of the image with replacement sky data corresponding to the applied image filter.

Abstract: The present disclosure provides a method, device and non-transitory computer storage medium for processing an image. The method includes: obtaining an image by a monocular camera; extracting image features with different levels based on the image; determining a fused feature by fusing the image features with different levels; and determining a depth distribution feature map of the image based on the fused feature, where a pixel value of each pixel point in the depth distribution feature map is a depth value.

Abstract: An active depth detection system can generate a depth map from an image and user interaction data, such as a pair of clicks. The active depth detection system can be implemented as a recurrent neural network that can receive the user interaction data as runtime inputs after training. The active depth detection system can store the generated depth map for further processing, such as image manipulation or real-world object detection.