Latest news with #DALL-E3
Fibre2Fashion
24-07-2025
- Fibre2Fashion
Generative AI to revolutionise fashion design: Research
Generative artificial intelligence (AI) has the potential to revolutionise fashion design. By recognising patterns in data and generating new text and images, AI models powered by deep learning algorithms can help fashion designers develop new catalogues, expanding creativity and helping bring products to market faster, as per a study by Pusan National University. Large language models (LLMs) like ChatGPT and AI image generators like DALL-E have shown promising results across industries and popularised the use of AI. In fashion, LLMs can help designers and non-experts understand past styles and predict future fashion trends. These insights can then generate prompts for AI image generators to produce real fashion collections. As such, it is increasingly important to understand how AI can be effectively integrated into fashion. In a recent study, professor Yoon Kyung Lee and master's student Chaehi Ryu, from the Department of Clothing and Textiles at Pusan National University, South Korea, explored how generative AI can contribute to visualising seasonal fashion trends. "To use AI effectively in fashion, we must understand the characteristics of generative AI models and make informed judgements of where they can be applied," explained. Lee. "In this study, we studied how effective prompt engineering can be used to generate realistic fashion collection images through AI." A study by Pusan National University shows that generative AI, using tools like ChatGPT and DALL-E 3, can help visualise and predict fashion trends. By analysing past data and crafting precise prompts, AI generated realistic Fall/Winter 2024 men's fashion images. While effective, limitations remain, highlighting the need for expert input. Using ChatGPT-3.5 and ChatGPT-4, the researchers first analysed men's fashion trends, based on historical data up to September 2021. From this, they used ChatGPT to predict men's fashion trends for Fall/Winter 2024. Design elements from these predictions were classified as 'initial codes'. In addition, design elements from Vogue's 2024 Fall/Winter Men's Fashion Trend data were used as 'modified codes', and those from literature as 'codes from literature'. These were then regrouped into six final codes: trends, silhouette elements, materials, key items, garment details, and embellishments. Using these codes, they created 35 prompts for DALL-E 3, each describing a unique outfit. The prompts followed a consistent template featuring a male model walking down a runway at a 2024 Fall/Winter fashion show. The template allowed customisation of event details, including aspect ratios, events, camera angles, model appearance and height, runway design, background, and audience details, and moods. Each prompt was run three times, generating a total of 105 images. DALL-E 3 was able to perfectly implement the prompts 67.6 per cent of the time. Prompts with adjectives demonstrated a high implementation rate. Some generated images closely resembled actual 2024 Fall/Winter Men's fashion collections. However, there were errors—most leaned toward ready-to-wear fashion, and DALL-E struggled with trend elements like gender fluidity. Trend keywords alone were insufficient to generate accurate results, indicating a need for further learning. "Our results show that expertly worded prompts are necessary for accurate fashion design implementation of generative AI, highlighting the important role of fashion experts," added Lee. "With further learning and improvements, generative AI models like DALL-E 3 will help fashion designers create entire fashion collections more efficiently, while supporting their creativity, and also help non-experts understand fashion trends." The study shows that generative AI can be a powerful tool not just for professionals but also for the general public, making it easier than ever to explore, predict, and style the upcoming season's fashion with confidence. Fibre2Fashion News Desk (RR)
Express Tribune
29-01-2025
- Business
- Express Tribune
DeepSeek launches Janus-Pro image generator, taking aim at DALL-E 3 superiority
Listen to article DeepSeek, the rapidly rising AI company, has unveiled a new suite of multimodal AI models under the Janus-Pro family, which it claims can outperform OpenAI's DALL-E 3. These models, ranging from 1 billion to 7 billion parameters in size, are available for download via the AI development platform Hugging Face. The Janus-Pro models are licensed under the MIT license, allowing for unrestricted commercial use. DeepSeek describes the Janus-Pro as a 'novel autoregressive framework' capable of both image analysis and creation. According to the company, the largest Janus-Pro model, Janus-Pro-7B, outperforms DALL-E 3 and other models like PixArt-alpha, Emu3-Gen, and Stability AI's Stable Diffusion XL on AI evaluation benchmarks such as GenEval and DPG-Bench. While some of these competing models are older, and most Janus-Pro models can only analyze smaller images (up to 384 x 384 resolution), the performance of Janus-Pro remains impressive considering its compact design. Photo: DeepSeek DeepSeek believes that Janus-Pro, with its simple yet powerful framework, surpasses previous unified models and challenges task-specific models. This makes it a strong contender in the field of next-generation unified multimodal models. The company has gained widespread attention after its chatbot app rose to the top of the Apple App Store charts. Funded primarily by High-Flyer Capital Management, a quantitative trading firm, DeepSeek's language models, developed using compute-efficient methods, have raised questions about the future of AI development and whether other nations can challenge U.S. dominance in the AI sector, especially regarding AI chip demand.
Yahoo
27-01-2025
- Business
- Yahoo
DeepSeek's Janus-Pro-7B AI Launch Causes Nvidia Stock to Drop 17%
On Monday, Chinese artificial intelligence startup DeepSeek revealed its newest multimodal AI model, Janus-Pro-7B, which set off a notable market response. While other AI-related equities, like Microsoft Corp. (MSFT, Financials), also sank over investor worries about growing expenses in the competitive AI field, Nvidia Corp. (NVDA, Financials) saw its shares plummet 17 percent. Warning! GuruFocus has detected 4 Warning Signs with NVDA. Designed to shine in knowledge and task creation, Janus-Pro-7B represents a significant development in artificial intelligence. Available on Hugging Face, the model makes use of an autoregressive framework and a consistent transformer architecture with independent visual encoding channels. This architecture lets Janus-Pro-7B surpass earlier unified models and challenge straight-forward specialized models like OpenAI's DALL-E 3. The model ranks well on main app stores and connects with DeepSeek's AI helper. High demand means that registration is only for Chinese phone numbers for now. Although Janus-Pro-7B is open-sourced under the MIT License, DeepSeek Model License governs use; developers may access and help to contribute to its repository on GitHub. Technical tools for Janus-Pro-7B include a quick-start tutorial accessible on Hugging Face and GitHub along with thorough documentation. The model uses the SigLIP-L vision encoder, competent of processing 384 by 384-pixel pictures, and has a downsample rate of 16 for image creation. This article first appeared on GuruFocus. Sign in to access your portfolio
Cedar News
27-01-2025
- Business
- Cedar News
DEEPSEEK DEBUTS OPEN-SOURCE AI MODEL JANUS-PRO-7B
Join our Telegram Chinese AI firm DeepSeek unveiled on Monday its open-source AI model, Janus-Pro-7B, which allegedly surpasses both Stable Diffusion and OpenAI's DALL-E 3 in terms of image creation capabilities. 'Janus-Pro is a novel autoregressive framework that unifies multimodal understanding and generation. It addresses the limitations of previous approaches by decoupling visual encoding into separate pathways, while still utilizing a single, unified transformer architecture for processing,' the company said on Monday. 'Janus-Pro surpasses previous unified model and matches or exceeds the performance of task-specific models
