Academia.edu no longer supports Internet Explorer.
To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to upgrade your browser.
2024, Journal of Ttechnology
In an era defined by the ubiquity of digital media, the demand for high-quality visuals has surged across various domains, from marketing and design to education and entertainment. However, creating these visuals often necessitates specialized skills and tools, limiting accessibility and inhibiting the creative potential of many individuals and professionals. Moreover, the rapid proliferation of misinformation and manipulated visuals underscores the importance of democratizing the image generation process while ensuring its reliability. The ever-growing presence of digital media has ignited a surge in demand for high-quality visuals across diverse fields, encompassing marketing, design, education, and entertainment. However, the creation of such visuals often necessitates specialized skills and intricate tools, hindering accessibility and obstructing the creative potential of many individuals and professionals. The proposed system addresses these challenges by leveraging cutting-edge deep learning techniques to develop an intuitive and user-centric system. This system empowers users to effortlessly generate images that directly correspond to their textual descriptions. By offering a solution that democratizes visual content creation, the approach not only tackles current accessibility limitations but also harbors the potential to bolster the authenticity and credibility of visual information within our increasingly image-driven digital landscape. This study introduces a novel system for text-to-image synthesis, enabling users to generate images corresponding to textual prompts. Leveraging advanced deep learning techniques, the system employs stateof-the-art generative models to bridge the gap between text and visual content. Users can input textual descriptions, keywords, or prompts, and the system translates these inputs into visually coherent and contextually relevant images. The approach aims to empower creative expression, assist content creators, and find applications in diverse domains such as art, design, and multimedia production. Through rigorous experimentation and evaluation, the study demonstrates the efficacy and versatility of the proposed text-driven image generation system, providing a valuable tool for harnessing the creative potential of human-AI collaboration.
xCoAx 2023. 11th Conference on Computation, Communication, Aesthetics & X
Creative Amplifiers: Augmenting Human Creativity with Text-to-Image Generators2023 •
This paper examines the extent to which deep-learning-based generative programs, particularly text-to-image generators, support human creativity in the sense Margaret Boden's definition. This discussion is supported by a brief introduction to the technical workings of denoising diffusion-based text-to-image generators. The analysis reveals that while these networks lack the autonomous ability to evaluate their designs and conduct exploratory design processes, they can nonetheless be considered complex tools that support human creativity by offering both accessible and powerful means of text-toimage translations. The paper then broadens its focus to a more general discussion of the potential impact of such assistance on creative labor, particularly in the design disciplines. Finally, the paper identifies the democratization of creativity as a larger disruptive force for creative labor than automatization, as professional workers might soon be competing with a larger, less trained workforce.
26th International Academic Mindtrek Conference
Perceptions and Realities of Text-to-Image GenerationText-to-image generation is a field with great potential. It is particularly interesting because it algorithmically synthesizes one data type into another, generating a photo-realistic image based off a phrase. One field of application is to aid language learning since children understand the meaning of a phrase better if they see images that correspond to text phrases. In this project, after exploration of several models, the Attentional Generative Adversarial Network model(AttnGAN) was found to have the best performance. The model was trained and validated via Google captions data-set. The LSTM was replaced with a Bidirectional Encoder Representations from Transformers, the parameters fine-tuned, and applied the resulting model is applied to the learning scenario where the quality of images generated from text captions is tested via several accuracy metrics.
ITM Web of Conferences
Successive Image Generation from a Single Sentence2021 •
Through various examples in history such as the early man’s carving on caves, dependence on diagrammatic representations, the immense popularity of comic books we have seen that vision has a higher reach in communication than written words. In this paper, we analyse and propose a new task of transfer of information from text to image synthesis. Through this paper we aim to generate a story from a single sentence and convert our generated story into a sequence of images. We plan to use state of the art technology to implement this task. With the advent of Generative Adversarial Networks text to image synthesis have found a new awakening. We plan to take this task a step further, in order to automate the entire process. Our system generates a multi-lined story given a single sentence using a deep neural network. This story is then fed into our networks of multiple stage GANs inorder to produce a photorealistic image sequence.
2019 •
Creating an image reflecting the content of a long text is a complex process that requires a sense of creativity. For example, creating a book cover or a movie poster based on their summary or a food image based on its recipe. In this paper we present the new task of generating images from long text that does not describe the visual content of the image directly. For this, we build a system for generating high-resolution 256 $\times$ 256 images of food conditioned on their recipes. The relation between the recipe text (without its title) to the visual content of the image is vague, and the textual structure of recipes is complex, consisting of two sections (ingredients and instructions) both containing multiple sentences. We used the recipe1M dataset to train and evaluate our model that is based on a the StackGAN-v2 architecture.
2019 IEEE/CVF International Conference on Computer Vision (ICCV)
Tell, Draw, and Repeat: Generating and Modifying Images Based on Continual Linguistic Instruction2019 •
2020 •
Semantic Image Generation refers to the task of generating photorealistic images conditioning on some input data. This task is carried out by a specific set of neural networks called Generative Adversarial Networks. These are a set of neural networks which work opposed to each other, evolving from each other’s successes. The generator is a convolution network that outputs some image, while the discriminator is a network that classifies said image. The job of the discriminator is to perfectly identify an image as fake or real, while the generator’s job is to try to produce realistic images. Combining Convolutional Neural Networks with a technique called Spatially Adaptive Normalization (which is similar to Batch Normalization), the results of semantic image generation tend to be less washed out, owing to the semantic information being “retained” throughout the network. Stylistic transfer is added by using Variational Auto Encoders, which take as input an image and breaks it down into...
One of the core applications of conditional generative models is to generate images from text (natural language). In addition to running tests on our capabilities of conditional modeling, dimensional distribution at the high level; synthesis involving text to an image has numerous practical application where most of them are exciting. Some of the applications include the creation of machine-aided content and editing photos. In the recent past, huge strides of progress in generative adversarial neural nets have been made. Text to image synthesis is among one of the most exciting discoveries made in the artificial intelligence field of our century. In the year 2016, the generative adversarial network for text to image synthesis was not able to property generate quite accurate and clear images. With the advancement made in technology and adjustments to the model, it’s now possible to generate clear and almost fully accurate images based on the description provided. “Its history in the making”. In this research document, we shall start with a brief introduction to the topic and some brief explanation of the workings of the model. Visualizing a scene given a detailed description of the scene in question is an undertaking that human beings do with less effort involved; nonetheless, it is a complex task that requires binding some abstract concepts that are described in words to how they look in real life. In this work we study previous work on image synthesis from text descriptions following the advances in generative adversarial networks (GANs), we go a step further and experiment with better training techniques like feature matching, smooth labeling, and mini-batch discrimination
Bartolomeo Bezzi (1851-1923). Un protagonista dimenticato dell'arte italiana tra Ottocento e Novecento - giornata di studi - Fucine di Ossana, 26 agosto 2023
Incontri improbabili: Bartolomeo Bezzi, Andrea Mantegna e Francesco Guardi2020 •
Ciencia Florestal
Ecologia da comunidade arbórea de Cerrado <i>stricto sensu</i> às margens de rodovias2023 •
2019 •
Case reports in psychiatry
Validity of Psychiatric Evaluation of Asylum Seekers through Telephone2021 •
Journal of microbiology, biotechnology and food sciences
The Impact of Asparaginase on Textural Properties of Wholegrain Cereal Biscuits Enriched with Sea Buckthorn PomaceLeadership and Management
Operational Analysis and Effective Resolution of Conflicts in the Sphere of Management2015 •
2009 •
Patient Preference and Adherence
The Impact of Social Media on the Acceptance of the COVID-19 Vaccine: A Cross-Sectional Study from Saudi Arabia2021 •