Mehmet Saygın Seyfioğlu

Currently @AmazonScience, phd @UW, previously @Google.

[Mehmet Saygın Seyfioğlu @mehsaygin]

Thank you @_akhaliq for sharing our work!

Check out Diffuse2Choose, a virtual try-all model that allows users to try on any e-commerce item in any setting!
Arxiv: https://t.co/zHHSOoZRtP
Website: https://t.co/qnMek6WqEy
Thread: https://t.co/GyjnVO2V8z

Stay tuned for more!

[AK @_akhaliq]

Amazon presents Diffuse to Choose

Enriching Image Conditioned Inpainting in Latent Diffusion Models for Virtual Try-All

paper page: https://t.co/jvhv9mPCue

Diffuse to Choose (DTC) allows users to virtually place any e-commerce item in any setting, ensuring detailed, semantically coherent blending with realistic lighting and shadows.

[Mehmet Saygın Seyfioğlu @mehsaygin]

Super excited to host the 2nd Evaluation of Generative Foundation Models
workshop at CVPR 2026!

Working on evaluating generative models? Submit your work to our CVPR 2026 Workshop!

📑Deadline: March 21 2026
📄 More details: https://t.co/xftUuxaw08

[Peyman Milanfar @docmilanfar]

how diffusion models work

[Mehmet Saygın Seyfioğlu @mehsaygin]

Check it out today at https://t.co/k6AHSTZwCv
For developers interested in integrating Nova Canvas VTO into your workflows, here's the technical blog: https://t.co/ILOzXw3REP

[Mehmet Saygın Seyfioğlu @mehsaygin]

Just over 15 months ago, this was nothing more than my research internship project at @AmazonScience. Today, I'm thrilled to announce we've officially launched Nova Canvas VTO — and I couldn't be more proud of how far we've come!
https://t.co/S93xX8PZWF via @YouTube

[Mehmet Saygın Seyfioğlu @mehsaygin]

What makes me most excited is that we've built something beyond traditional virtual try-on technology. While most solutions only let you try on garments, we've created what we call a "virtual try-all" experience. Now you can visualize accessories, furniture, or virtually anything you can think of.

[Mehmet Saygın Seyfioğlu @mehsaygin]

@natolambert Probably most of it are spent on reasoning.

[Mehmet Saygın Seyfioğlu @mehsaygin]

@yigitokar @Google Care-ai looks neat btw. But needs free tier for us peasants.

[Mehmet Saygın Seyfioğlu @mehsaygin]

Thanks for the profile pic guys @Google

[Mehmet Saygın Seyfioğlu @mehsaygin]

Check out Pathfinder! We’ve imitated human doctors’ diagnostic process in whole slide images using a multi-agent framework. Pathfinder not only surpasses the state of the art AI models in skin cancer detection but also outperforms human experts! Code, data and model coming soon!

[Fatemeh Ghezloo @fghezloo]

🤖🩺 Introducing PathFinder: A multi-agent system that navigates WSIs akin to pathologists, collecting findings towards interpretable diagnosis. It achieves 74% accuracy on melanoma diagnosis, outperforming experts by 9%!

🌎: https://t.co/s6yLlLGJ7n
📰: https://t.co/1GWdxuxbyN

[Mehmet Saygın Seyfioğlu @mehsaygin]

@Yuchenj_UW

[Rohan Paul @rohanpaul_ai]

How much GPU VRAM do I need to be able to serve Llama 70B?

Clean formulae

[Mehmet Saygın Seyfioğlu @mehsaygin]

Just presented Quilt-LLaVA @CVPR with @wizdom_dominic and @fghezloo! Huge thanks to everyone who stopped by to chat with us!

[Mehmet Saygın Seyfioğlu @mehsaygin]

Excited to present Quilt-LLaVA https://t.co/Snm8ZzeBH6 with my co-leads @wizdom_dominic and @fghezloo @CVPR tomorrow! Join us at the poster session from 10 AM to 12 PM in ARCH 4A-E #342

[Mehmet Saygın Seyfioğlu @mehsaygin]

We are organizing "The First Workshop on the Evaluation of Generative Foundation Models" @CVPR next Tuesday (Summit 433, 1-6:30 pm).

For details, please see https://t.co/1q7D6yTcxO

See you all in Seattle! #CVPR24

[Mehmet Saygın Seyfioğlu @mehsaygin]

Excited to be giving an invited talk about https://t.co/WSkj21njSb at "The First Workshop on Virtual Try-On" @CVPR (Monday 4:40 pm, Room Arch 309)!

Details: https://t.co/fFCwuYSJqZ

Looking forward to seeing all those who attend in person!

[Mehmet Saygın Seyfioğlu @mehsaygin]

Excited to share that Quilt-LLaVA has been accepted at @CVPR! As promised, we are making our model, datasets, and training scripts available to the community: https://t.co/Snm8ZzeBH6

See you all in Seattle!
@wizdom_dominic @fghezloo @RanjayKrishna @lindashapiro

[Mehmet Saygın Seyfioğlu @mehsaygin]

Introducing Quilt-LLaVA, a Large Language and Vision Assistant for #Pathology trained with spatially localized instruction tuning data generated from educational #YouTube videos, outperforming SOTA in various tasks.

🌎: https://t.co/vqLYKD7SAx
📜: https://t.co/l5tvpo401U
A 🧵:

[Mehmet Saygın Seyfioğlu @mehsaygin]

Thank you, @MariaZontak, for doing the heavy lifting in organizing this workshop, and thanks to the rest of the team at @AmazonScience!

[Mehmet Saygın Seyfioğlu @mehsaygin]

Call for papers on the First Workshop on Evaluation for Generative Foundation Models @CVPR. We're eager to hear your ideas on how generative models should be evaluated!

We also have a stellar speaker lineup: @THassner, Bo Li, @RanjayKrishna, Hanwang Zhang, @leokarlin, @jungokasai, Sadeep Jayasumana, @besanushi and @ecekamar.

Submission Deadline: April 4, 23:59 PST

More details: https://t.co/ciZcgdkz9y

See you all in Seattle!

[Mehmet Saygın Seyfioğlu @mehsaygin]

@xiaohuggg @forrestbing Thank you @xiaohuggg , for covering our work! Please see the following thread for more information!
https://t.co/GyjnVO2V8z

Also, please consider following so you won't miss it when we share the codebase and the model 😅

[Mehmet Saygın Seyfioğlu @mehsaygin]

Excited to share our latest work from my
@AmazonScience internship: Diffuse2Choose, which is a zero-shot diffusion model that enables "Virtual Try-All" – the ability to virtually try on any e-commerce item!
🌎: https://t.co/TwT709uP2j
📜: https://t.co/zHHSOoZRtP
A 🧵: