Exploring the intersection of ๐๐๐๐๐ creativity ๐ฅ and ๐จ๐ฐ ๐ค to achieve new heights in design aesthetics. We're unlocking technology to reach this goal.
Joined July 2024
- Tweets32
- Following2
- Followers35
- Likes1
Proud that our HF port of GenPoster100k (org. SAGE paper) was featured in a recent AI compliance study!
Out of 11,682 datasets, ours was named an ๐๐๐๐๐๐๐๐๐๐๐๐๐ ๐๐๐๐, gold-standard example of a highly detailed Dataset Card.
๐งต Links to the papers & dataset below
๐ Links & Resources:
๐ The compliance study (Highlighting our Dataset Card):
arxiv.org/abs/2607.03310
๐๏ธ Our Hugging Face release (GenPoster100K):
huggingface.co/datasets/creaโฆ
๐ The original dataset paper (SAGE):
arxiv.org/abs/2510.15749
What if we applied the concept ๐พ๐๐๐๐๐กโ๐ฆ'๐ ๐ฟ๐ฟ๐ ๐ค๐๐๐ to creative graphic design?
Weโre building ๐พ๐ง๐๐๐ฉ๐๐ซ๐ ๐๐ง๐๐ฅ๐๐๐ ๐ฟ๐๐จ๐๐๐ฃ ๐ฟ๐๐ฉ๐๐จ๐๐ฉ๐จ, a growing collection of 20+ Hugging Face๐ค-ready datasets for graphic design generation and evaluation.
From posters and banners to UIs, typography, layouts, ads, evaluation, and more.
Link below ๐
Released a FastAPI server for image layer decomposition using Qwen-Image-Layered!
โจ๏ธ Automatically decompose images into semantic layers
โก๏ธ FastAPI + GPU acceleration
๐ ๏ธ Simple REST API
GitHub: github.com/creative-graphic-โฆ
๐ฅ Want LongCLIP in a familiar API? Now you can use LongCLIP with a Transformers-compatible API โ load models like any other ๐ค Transformers model and run long-text vision-language inference with the `.from_pretrained` function.
โ After & Before โ
๐ Link in the next tweet ๐
๐ Just dropped: gpt-graphic-design-evaluator โ a LangChain-powered system to automatically evaluate graphic design via LLMs!
๐ Based on the paper โCan GPTs Evaluate Graphic Design Based on Design Principles?โ arxiv.org/abs/2410.08885
๐ก Check it out: github.com/creative-graphic-โฆ
๐ญ Iโve seen some researchers who have been working on creative graphic design research recently move to Canva ๐ง
> F Lite is a 10B parameter diffusion model created by Freepik and Fal, trained exclusively on copyright-safe and SFW content. The model was trained on Freepik's internal dataset comprising approximately 80 million copyright-safe images
huggingface.co/Freepik/F-Litโฆ
๐ Introducing lighting-diffusers: your template for diffusion model training with ๐คDiffusers & โกPyTorch Lightning!
With full LightningCLI support, load all experiment settings from YAML filesโmaking workflows reproducible, transparent, and manageable.
github.com/creative-graphic-โฆ
common-canvas/commoncatalog-cc-by ยท Datasets at Hugging Face huggingface.co/datasets/commโฆ
nyuuzyou/publicdomainpictures ยท Datasets at Hugging Face huggingface.co/datasets/nyuuโฆ
Thrilled to share our published #SSII2024 technical map, showcasing the evolution of layout and poster generation research from 1990 to 2024! Excited to see how the field evolves with future directions in creative design๐
๐บ๏ธ bit.ly/ssii2024-techmap-creaโฆ
๐ bit.ly/ssii2024-techmap-creaโฆ
๐ฃ Our partner has released ๐ช๐ ๐๐ฎ๐ค-๐-๐๐๐๐-๐๐ก๐๐ฅ๐ฎ-๐๐๐ฅ๐ฉ๐๐ค๐ฃ๐จ dataset, which is a Ukiyo-e dataset with captions added using BLIP2. We intend to use this dataset in LoRA tutorials and other such materials:
๐ค huggingface.co/datasets/py-iโฆ
๐ github.com/py-img-gen/hugginโฆ
laion/relaion2B-en-research ยท Datasets at Hugging Face huggingface.co/datasets/laioโฆ
laion/relaion2B-en-research-safe ยท Datasets at Hugging Face huggingface.co/datasets/laioโฆ
