-
公开(公告)号：GB2623401B
公开(公告)日：2025-01-08
申请号：GB202311866
申请日：2023-08-02
Applicant: ADOBE INC
Inventor： ZHE LIN , HAITIAN ZHENG , ELYA SHECHTMAN , JIANMING ZHANG , JINGWAN LU , NING XU , QING LIU , SCOTT COHEN , SOHRAB AMIRGHODSI
Abstract: The present disclosure relates to systems, methods, and non-transitory computer readable media for panoptically guiding digital image inpainting utilizing a panoptic inpainting neural network. In some embodiments, the disclosed systems utilize a panoptic inpainting neural network to generate an inpainted digital image according to panoptic segmentation map that defines pixel regions corresponding to different panoptic labels. In some cases, the disclosed systems train a neural network utilizing a semantic discriminator that facilitates generation of digital images that are realistic while also conforming to a semantic segmentation. The disclosed systems generate and provide a panoptic inpainting interface to facilitate user interaction for inpainting digital images. In certain embodiments, the disclosed systems iteratively update an inpainted digital image based on changes to a panoptic segmentation map.
-
12.发明专利
公开(公告)号：GB2587841A
公开(公告)日：2021-04-14
申请号：GB202007986
申请日：2020-05-28
Applicant: ADOBE INC
Inventor： ZHE LIN , JIANMING ZHANG , HE ZHANG , FEDERICO PERAZZI
Abstract: The present disclosure relates to utilizing a neural network having a two-stream encoder architecture to accurately generate composite digital images that realistically portray a foreground object from one digital image against a scene from another digital image. For example, the disclosed systems can utilize a foreground encoder of the neural network to identify features from a foreground image, 310 to generate a foreground segmentation mask, 314 and further utilize a background encoder to identify features from a background image, 312 by means of an inverted foreground segmentation mask, 316. The disclosed systems can then utilize a decoder, 300 to fuse the features together and generate a composite digital image, 318. The disclosed systems can train the neural network utilizing an easy-to-hard data augmentation scheme implemented via self-teaching. The disclosed systems can further incorporate the neural network within an end-to-end framework for automation of the image composition process.
-
公开(公告)号：GB2567920B
公开(公告)日：2021-02-17
申请号：GB201813276
申请日：2018-08-15
Applicant: ADOBE INC
Inventor： ZHE LIN , XIN LU , XIAOHUI SHEN , JIMEI YANG , JIANMING ZHANG , JEN-CHAN JEFF CHIEN , CHENXI LIU
Abstract: Systems, methods, and non-transitory computer-readable media are disclosed for segmenting objects in digital visual media utilizing one or more salient content neural networks. In particular, in one or more embodiments, the disclosed systems and methods train one or more salient content neural networks to efficiently identify foreground pixels in digital visual media. Moreover, in one or more embodiments, the disclosed systems and methods provide a trained salient content neural network to a mobile device, allowing the mobile device to directly select salient objects in digital visual media utilizing a trained neural network. Furthermore, in one or more embodiments, the disclosed systems and methods train and provide multiple salient content neural networks, such that mobile devices can identify objects in real-time digital visual media feeds (utilizing a first salient content neural network) and identify objects in static digital images (utilizing a second salient content neural network).
-
公开(公告)号：GB2567920A
公开(公告)日：2019-05-01
申请号：GB201813276
申请日：2018-08-15
Applicant: ADOBE INC
Inventor： ZHE LIN , XIN LU , XIAOHUI SHEN , JIMEI YANG , JIANMING ZHANG , JEN-CHAN JEFF CHIEN , CHENXI LIU
Abstract: Systems, methods, and computer-readable media are disclosed for segmenting objects in digital visual media utilizing one or potentially, more salient content neural networks. A mobile device, such as a mobile or cell phone or tablet 300, captures real-time images 302 and applies a real-time neural network 304 embedded in the mobile device, to train a network to predict foreground pixels from a captured training image, comparing the predicted images with ground truth pictures. The trained network is then used identify objects in real-time digital media 306 and generate a modified image for display, based on said identified features 308. The digital image may be part of real-time digital media feed. A second neural network may be present on the mobile device, using training data which is different from the first training image. Background pixels may be separated from foreground ones, and both background and foreground pixels may be modified. A system where the mobile device acts as a client, while the neural network operates on the server, is also envisaged (Figure 1).
-
15.发明专利
公开(公告)号：GB2628694A
公开(公告)日：2024-10-02
申请号：GB202319660
申请日：2023-12-20
Applicant: ADOBE INC
Inventor： KRISHNA KUMAR SINGH , YIJUN LI , JINGWAN LU , DUYGU CEYLAN AKSIT , YANGTUANFENG WANG , JIMEI YANG , TOBIAS HINZ , QING LIU , JIANMING ZHANG , ZHE LIN
Abstract: The present disclosure relates to systems, methods, and computer-readable media that modify digital images via scene-based editing using image understanding facilitated by artificial intelligence. One or more embodiments the disclosed systems utilize generative machine learning models to create modified digital images portraying human subjects. The invention comprises determining a region of a human portrayed within a digital image to inpaint and generating, utilising a generative segmentation machine learning model, an infill segmentation map from the digital image where the infill segmentation map comprising a human segmentation classification for the region. Utilizing a human inpainting generative adversarial neural network, a modified digital image from the digital image and the infill segmentation map a modified image is generated comprising modified pixels for the region corresponding to the human segmentation classification.
-
Patent Agency Ranking
公开(公告)号：GB2628691A
公开(公告)日：2024-10-02
申请号：GB202318853
申请日：2023-12-11
Applicant: ADOBE INC
Inventor： KRISHNA KUMAR SINGH , YIJUN LI , JINGWAN LU , DUYGU CEYLAN AKSIT , YANGTUANFENG WANG , JIMEI YANG , TOBIAS HINZ , QING LIU , JIANMING ZHANG , ZHE LIN
Abstract: The present disclosure relates to a system, method, and computer-readable media that modify digital images. In particular a depiction of a human and a region of the human to inpaint is determined from a digital image 5504 and, using an encoder, a structural encoding 5508 from a structure guidance map of the human is generated. Using an encoder, a visual appearance encoding 5510 from the human portrayed in the digital image is also generated. Utilizing a human inpainting generative adversarial neural network 5514, a modified digital image 5518 comprising modified pixels of the region is generated from the structural encoding and the visual appearance encoding of the human. Furthermore, a system generates a local appearance feature tensor 5516 using a parameter neural network 5512 from the visual appearance encoding. The system may generate a spatially varying scaling tensor and a spatially varying shifting tensor by utilizing the parameter neural network.
-
公开(公告)号：GB2560786B
公开(公告)日：2021-09-22
申请号：GB201720901
申请日：2017-12-14
Applicant: ADOBE INC
Inventor： ZHE LIN , XIAOHUI SHEN , JIANMING ZHANG , HAILIN JIN , YINGWEI LI
Abstract: A framework is provided for associating dense images with topics. The framework is trained utilizing images, each having multiple regions, multiple visual characteristics and multiple keyword tags associated therewith. For each region of each image, visual features are computed from the visual characteristics utilizing a convolutional neural network, and an image feature vector is generated from the visual features. The keyword tags are utilized to generate a weighted word vector for each image by calculating a weighted average of word vector representations representing keyword tags associated with the image. The image feature vector and the weighted word vector are aligned in a common embedding space and a heat map is computed for the image. Once trained, the framework can be utilized to automatically tag images and rank the relevance of images with respect to queried keywords based upon associated heat maps.
-
公开(公告)号：GB2573849A
公开(公告)日：2019-11-20
申请号：GB201901759
申请日：2019-02-08
Applicant: ADOBE INC
Inventor： ZHE LIN , XIAOHUI SHEN , MINGYANG LING , JIANMING ZHANG , JASON KUEN , BRETT BUTTERFIELD
IPC: G06F16/532 , G06F16/583
Abstract: A computer readable medium for matching digital images based on visual similarity comprises instructions that cause a computer device to receive, from a user client device, a user selection of a query digital image and at least one of spatial selectivity, image composition, or object count to use to identify similar digital images, utilise a deep neural network-based model to generate a set of deep features for the query digital image, generate, based on the set of deep features of the query digital image and in accordance with the user selection, a deep neural network-based representation of the query digital image by utilising one or more of a spatial selectivity algorithm, an image composition algorithm, or an object count algorithm, and based on the deep neural network-based representation of the query digital image, identify, from a digital image database, a similar digital image for the query digital image, for example by determining similarity scores and ranking the images in the database. An image mask may be received by the user, which indicates a portion of the query image to emphasise in identifying similar images, by weighting the portion of the query image indicated by the mask.
-
公开(公告)号：GB2568118A
公开(公告)日：2019-05-08
申请号：GB201720898
申请日：2017-12-14
Applicant: ADOBE INC
Inventor： ZHE LIN , XIAOHUI SHEN , JIANMING ZHANG , HAILIN JIN , YINGWEI LI
IPC: G06F16/58
Abstract: A method of calculating the relevance of tags applied to images involves receiving an image 314 with associated tags, which may be user-applied tags. The tags are used to create a weighted word vector 218, also known as a soft topic vector, which represents the dominant concept among the keyword tags. Visual features 312 of the image may be used to create an image feature vector 310 which can then be aligned in a common embedding space. The aligned vectors can then be used to calculate a relevancy score for each tag as it pertains to the image. The visual features may be determined with a convolutional neural network. The weighted word and image feature vectors may be aligned using cosine similarity loss. Each tag may be assigned a word vector representation and a weighted average of the word vectors can then be used to generate the weighted word vector.
-
公开(公告)号：US20200250465A1
公开(公告)日：2020-08-06
申请号：US16853111
申请日：2020-04-20
Applicant: Adobe Inc.
Inventor： ZHE LIN , XIAOHUI SHEN , JONATHAN BRANDT , JIANMING ZHANG , CHEN FANG
IPC: G06K9/62 , G06F16/51 , G06F16/28 , G06F16/2457 , G06K9/46 , G06F16/583 , G06N3/04 , G06N3/08
Abstract: Embodiments of the present invention provide an automated image tagging system that can predict a set of tags, along with relevance scores, that can be used for keyword-based image retrieval, image tag proposal, and image tag auto-completion based on user input. Initially, during training, a clustering technique is utilized to reduce cluster imbalance in the data that is input into a convolutional neural network (CNN) for training feature data. In embodiments, the clustering technique can also be utilized to compute data point similarity that can be utilized for tag propagation (to tag untagged images). During testing, a diversity based voting framework is utilized to overcome user tagging biases. In some embodiments, bigram re-weighting can down-weight a keyword that is likely to be part of a bigram based on a predicted tag set.
-
-
-
-
-
-
-
-
-
Search Results
25
records
(took 0.021 seconds)
