| former: RS on GenAI for graphic design , Ph.D. in UTokyo , intern , eeic2014, tk60, 
Type-R retouches erroneous text renderings by recent text-to-image models via the post-processing pipeline, achieving a much better tradeoff between image quality and text accuracy.
arxiv.org/abs/2411.18159
Happy to announce our #CVPR2023 work!
(please RT if you like)
"LayoutDM: Discrete Diffusion Model for Controllable Layout Generation"
[paper] arxiv.org/abs/2303.08137
[proj.] cyberagentailab.github.io/layout-dm/
We solve various conditional layout generation tasks using a single model.
Our #CVPR2023 (highlight) work!
"Towards Flexible Multi-modal Document Models"
We help users generate graphic designs in a vector format by multi-modal fill-in-the-blank.
[paper] arxiv.org/abs/2303.18248
[proj.] cyberagentailab.github.io/flex-dm/
arxiv.org/abs/2311.13602
udonda.github.io/RALF/
Happy to announce our latest preprint!
"OpenCOLE: Towards Reproducible Automatic Graphic Design Generation"
[paper] arxiv.org/abs/2406.08232
[proj.] github.com/CyberAgentAILa
We propose an open framework for automatic graphic design and release the model trained on open data.
ちなみに前職の時の仕事をしゃべるワークショップ招待講演もあります
（質問もDMその他でお気軽にどうぞ）ほかのチームでもCV系のトピック扱ってますのでぜひ一度ご覧ください〜
Happy to announce our #CVPR2023 work!
(please RT if you like)
"LayoutDM: Discrete Diffusion Model for Controllable Layout Generation"
[paper] arxiv.org/abs/2303.08137
[proj.] cyberagentailab.github.io/layout-dm/
We solve various conditional layout generation tasks using a single model.
