786 Hits in 3.7 sec

Toward a Visual Concept Vocabulary for GAN Latent Space [article]

Sarah Schwettmann, Evan Hernandez, David Bau, Samuel Klein, Jacob Andreas, Antonio Torralba
2021 arXiv   pre-print
This paper introduces a new method for building open-ended vocabularies of primitive visual concepts represented in a GAN's latent space.  ...  A large body of recent work has identified transformations in the latent spaces of generative adversarial networks (GANs) that consistently and interpretably transform generated images.  ...  We thank the MIT-IBM Watson AI Lab for support, and IBM for the donation of the Satori supercomputer that enabled training BigGAN on MIT Places.  ... 
arXiv:2110.04292v1 fatcat:lvfbx47ssrej7nlrtasgzckuh4

Towards Highly Expressive Machine Learning Models of Non-Melanoma Skin Cancer [article]

Simon M. Thomas, James G. Lefevre, Glenn Baxter, Nicholas A.Hamilton
2022 arXiv   pre-print
Pathologists have a rich vocabulary with which they can describe all the nuances of cellular morphology. In their world, there is a natural pairing of images and words.  ...  Implementing a VQ-GAN model to reconstruct high-resolution (256x256) images of IEC images, we trained a sequence-to-sequence transformer to generate natural language descriptions using pathologist terminology  ...  Acknowledgements We wish to acknowledge The University of Queensland's Research Computing Centre (RCC) for its support in this research.  ... 
arXiv:2207.05749v1 fatcat:xdwsh5xndvhn3esbd4yzy6dbmi

Feature Learning with Adversarial Networks for Concept Detection in Medical Images: UA.PT Bioinformatics at ImageCLEF 2018

Eduardo Pinho, Carlos Costa
2018 Conference and Labs of the Evaluation Forum  
Subsequently, two kinds of classification algorithms were employed for concept detection over the feature spaces learned.  ...  This paper describes a set of feature learning approaches for the concept detection sub-task of ImageCLEFcaption 2018.  ...  Funds through the FCT -Fundação para a Ciência e a Tecnologia within project PTDC/EEI-ESS/6815/2014.  ... 
dblp:conf/clef/Pinho018 fatcat:vvfcsuhhqfbslnjipqhgcfzucq

Unsupervised Learning for Concept Detection in Medical Images: A Comparative Analysis

Eduardo Pinho, Carlos Costa
2018 Applied Sciences  
Each model was trained, and their respective feature spaces evaluated using images from the ImageCLEF 2017 concept detection task.  ...  In this paper, we present an assessment of unsupervised feature learning approaches for images in biomedical literature which can be applied to automatic biomedical concept detection.  ...  Funds through the FCT -Fundação para a Ciência e a Tecnologia, within project PTDC/EEI-ESS/6815/2014.  ... 
doi:10.3390/app8081213 fatcat:ok6kpyilp5gqda6w7k2gkqjg6a

Ideation via Critic-Based Exploration of Generator Latent Space

Puneet Jain, Najma Mathema, Jonathan Skaggs, Dan Ventura
2021 International Conference on Computational Creativity  
This process is iterated by using feedback from the evaluation module to guide exploration of the latent space of the GAN.  ...  We present a system for generating, evaluating, and refining logos that can act as a collaborator for creating relevant logo designs.  ...  feedback to further explore the GAN's latent space. concepts visually (2018).  ... 
dblp:conf/icccrea/JainMSV21 fatcat:3k4d6bl6yzcbnlbfvgqto6liba

Fashion Style Generation: Evolutionary Search with Gaussian Mixture Models in the Latent Space [article]

Imke Grabe, Jichen Zhu, Manex Agirrezabal
2022 arXiv   pre-print
Finding the latent vectors in the generator's latent space that correspond to a style is approached as an evolutionary search problem.  ...  Showing that the developed system can generate images of maximum fitness visually resembling certain styles, our approach provides a promising direction to guide the search for style-coherent designs.  ...  Evolutionary search of GANs' latent space Our project aims to generate designs of various styles with visually diverse attributes with GANs.  ... 
arXiv:2204.00592v2 fatcat:bmoxmyzseva4jperyii3eegrve

Paint by Word [article]

David Bau, Alex Andonian, Audrey Cui, YeonHwan Park, Ali Jahanian, Aude Oliva, Antonio Torralba
2021 arXiv   pre-print
We find that, to make large changes, it is important to use non-gradient methods to explore latent space, and it is important to relax the computations of the GAN to target changes to a specific region  ...  be able to point to a location in a synthesized image and apply an arbitrary new concept such as "rustic" or "opulent" or "happy dog."  ...  We thank IBM for the donation to MIT of the Satori supercomputer that enabled training BigGAN on MIT Places, and we thank OpenAI, Google, and Nvidia for publishing weights for pretrained large-scale CLIP  ... 
arXiv:2103.10951v2 fatcat:55sefh6mk5fdpabutzbm7g2jfa

Open-Edit: Open-Domain Image Manipulation with Open-Vocabulary Instructions [article]

Xihui Liu, Zhe Lin, Jianming Zhang, Handong Zhao, Quan Tran, Xiaogang Wang, Hongsheng Li
2021 arXiv   pre-print
Our approach takes advantage of the unified visual-semantic embedding space pretrained on a general image-caption dataset, and manipulates the embedded visual features by applying text-guided vector arithmetic  ...  Our approach shows promising results in manipulating open-vocabulary color, texture, and high-level attributes for various scenarios of open-domain images.  ...  GAN-Paint [4] manipulates the latent space of the input image guided by GAN Dissection [5] , which relies on a segmentation model to identify latent units related to specific objects.  ... 
arXiv:2008.01576v2 fatcat:zrvf7mdo4zeuzcryu2jhmd7any

Towards a Framework for Human-AI Interaction Patterns in Co-Creative GAN Applications 92-102

Imke Grabe, Miguel González Duque, Sebastian Risi, Jichen Zhu
2022 International Conference on Intelligent User Interfaces  
While generative models have been applied in various creative tasks across disciplines, a theoretical foundation for understanding human-GAN collaboration is yet to be developed.  ...  Drawing from the mixed-initiative co-creation community, we propose a preliminary framework to analyze co-creative GAN applications.  ...  One direction to move an artifact through latent space is towards other artifacts in the along latent directions of generated a Mario level [21] , and along semantic features in CG-GAN [23] . space.  ... 
dblp:conf/iui/GrabeDRZ22 fatcat:jqbir7iyk5b7feul6querdrazy

HiGAN+: Handwriting Imitation GAN with Disentangled Representations

Ji Gan, Weiqiang Wang, Jiaxu Leng, Xinbo Gao
2022 ACM Transactions on Graphics  
GANs for handwritten text synthesis.  ...  Humans remain far better than machines at learning, where humans require fewer examples to learn new concepts and can use those concepts in richer ways.  ...  Despite the latent style space, we also perform the interpolation in the text space to further validate the generalization of HiGAN+.  ... 
doi:10.1145/3550070 fatcat:ws4v6dzckravrlkobmkabgwkre

GANwriting: Content-Conditioned Generation of Styled Handwritten Word Images [article]

Lei Kang, Pau Riba, Yaxing Wang, Marçal Rusiñol, Alicia Fornés, Mauricio Villegas
2020 arXiv   pre-print
Our model is unconstrained to any predefined vocabulary, being able to render whatever input word. Given a sample writer, it is also able to mimic its calligraphic features in a few-shot setup.  ...  Our generator is guided by three complementary learning objectives: to produce realistic images, to imitate a certain handwriting style and to convey a specific textual content.  ...  In the original GAN architecture, inputs were randomly sampled from a latent space, so that it was hard to control which kind of images were being generated.  ... 
arXiv:2003.02567v2 fatcat:ooztnppezraltfgdczankeryru

Automatic Concept Discovery from Parallel Text and Visual Corpora

Chen Sun, Chuang Gan, Ram Nevatia
2015 2015 IEEE International Conference on Computer Vision (ICCV)  
How to build a similar connection for computers? One possible way is via visual concepts, which are text terms that relate to visually discriminative entities.  ...  large sets of manually selected concepts significantly, but also achieves the stateof-the-art performance in the retrieval task. * This work was done when Chuang Gan was a visiting researcher at University  ...  Acknowledgement: We thank Kevin Knight for helpful discussions.  ... 
doi:10.1109/iccv.2015.298 dblp:conf/iccv/SunGN15 fatcat:n5viaszy7ne4zmiffsm7c2sizu

The Many Moods of Emotion [article]

Valentin Vielzeuf, Corentin Kervadec, Stéphane Pateux, Frédéric Jurie
2018 arXiv   pre-print
Building upon the assumption of the psychological community that emotion is intrinsically continuous, we first design our own continuous emotion representation with a 3-dimensional latent space issued  ...  Finally we show from visual interpretation, that the third remaining dimension is highly related to the well-known dominance dimension from psychology.  ...  We therefore propose to use a latent space of a convolutional neural network.  ... 
arXiv:1810.13197v1 fatcat:7pdufk4d4vanfi4byculayreua

Information Maximizing Visual Question Generation [article]

Ranjay Krishna, Michael Bernstein, Li Fei-Fei
2019 arXiv   pre-print
We launch our model on a set of real world images and extract previously unseen visual concepts.  ...  We regularize this latent space with a second latent space that ensures clustering of similar answers.  ...  We thank Justin Johnson, Andrey Kurenkov, Apoorva Dornadula and Vincent Chen for their helpful comments and edits.  ... 
arXiv:1903.11207v1 fatcat:sdcq2j5fvrhmhgsxgjaa7mmvaa

Information Maximizing Visual Question Generation

Ranjay Krishna, Michael Bernstein, Li Fei-Fei
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
We launch our model on a set of real world images and extract previously unseen visual concepts.  ...  We regularize this latent space with a second latent space that ensures clustering of similar answers.  ...  We thank Justin Johnson, Andrey Kurenkov, Apoorva Dornadula and Vincent Chen for their helpful comments and edits.  ... 
doi:10.1109/cvpr.2019.00211 dblp:conf/cvpr/KrishnaB019 fatcat:rosaexkgurgxliemdiidvi32di
« Previous Showing results 1 — 15 out of 786 results