Douglas Gray @de3ug X Profile

Douglas Gray

@de3ug

Followers

464

Following

11K

Media

244

Statuses

2K

Visual Search @amazon. Here for computer vision, ML, NLP research discussion, but may drift into hobbies like astrophotography.

https://t.co/o2s25iXB5G

Portola Valley, CA

Joined June 2009

Don't wanna be here? Send us removal request.

@levelsio

1 month

Business class was full, also it was a short flight and this is also fun I think it's more fun to mix very cheap and very luxury to feel the contrast I started to realize luxury business class, and luxury hotels, and all that, it just gets very boring very fast if you don't

ko₿.eth aka 0xko₿ΞOS

@CryptoKobe

1 month

@levelsio @CSAIRGlobal Bro how are you not wealthy enough to at least fly business class?

127

52

3K

Aditya Ramesh

@model_mechanic

1 month

The value of fast iteration in AI is overrated. The best results are obtained by knowing the right things to do and doing each thing with neurotic precision and attention to detail.

26

37

435

Yaroslav

@512x512

1 month

This bulge on a Waymo map seems funny. Which Google exec lives there?

107

98

11K

Douglas Gray

@de3ug

2 months

This is why I like fabric posters. Easy to fold up and bring home to decorate the lab.

Jia-Bin Huang

@jbhuang0604

2 months

graveyard of knowledge … RIP

0

1

Douglas Gray

@de3ug

2 months

Good reminder for why it’s important to attend the PAMI-TC meeting. If you don’t, well meaning people will vote for more bureaucratic requirements.

Jon Barron

@jon_barron

2 months

It looks like @CVPR has implemented a new mandatory "Compute Reporting Form" that must be submitted alongside any paper submission. Though I am sympathetic to the motivations for this change, I am opposed to it for a variety of reasons:

1

0

2

Douglas Gray

@de3ug

2 months

Huge thanks to everyone who made MRR @ ICCV possible — our authors, reviewers, attendees, and speakers: Roi Herzig, Cordelia Schmid, Jianwei Yang, Kristen Grauman, and the ICCV organizers. 🙏 See you next year! #ICCV2025 #AI #ComputerVision #Multimodal #Retrieval

0

Douglas Gray

@de3ug

2 months

All papers from the MRR Workshop @ #ICCV2025 are now live on CVF Open Access 📚 🔗

1

0

2

Douglas Gray

@de3ug

2 months

The paper doesn’t try rendering arbitrary text and feeding it through this system instead of tokens, but I agree with @karpathy that this is where things are headed. We’ll likely render text directly into image space and let multimodal models handle it end-to-end.

0

Douglas Gray

@de3ug

2 months

It doesn’t just speed up OCR—it shows how vision and language models can share a single representational layer. The visual and text transformers operate on the same logic, blurring the line between seeing and reading.

1

0

Douglas Gray

@de3ug

2 months

Lots of interest in this work, but many people seem to be missing the point or arguing over who did it first, which says a lot about its importance. DeepSeek-OCR is a big step because it’s a true multimodal model, not a case of attaching a Vision Transformer to an LLM.

Andrej Karpathy

@karpathy

2 months

I quite like the new DeepSeek-OCR paper. It's a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn't matter. The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language

2

0

2

Douglas Gray

@de3ug

2 months

MRR Keynote starting now — Kristen Grauman (11:45–12:25). Room 308B. #ICCV2025

0

1

Douglas Gray

@de3ug

2 months

Keynote at MRR starting now — Jianwei Yang (10:50–11:30). Room 308B. #ICCV2025

0

Douglas Gray

@de3ug

2 months

Up next at MRR: Keynote — Cordelia Schmid (8:55–9:35). Room 308B. #ICCV2025

0

1

Douglas Gray

@de3ug

2 months

Starting in 10 minutes. Grab your coffee and come by 308B to hear about multimodal representations & robots! #ICCV2025

Roei Herzig

@roeiherzig

2 months

Happening tomorrow at 8:35 AM — the Multimodal Representation & Retrieval (MRR) workshop at #ICCV2025! 🎉 Come hear how multimodality powers robot learning 🤖✨ 📍 Room 308 B

0

5

Douglas Gray

@de3ug

2 months

Lots of folks eating lunch on the floor. @ICCVConference should really get more chairs near the food. #ICCV2025

1

2

33

Roei Herzig

@roeiherzig

2 months

🌺Super excited to present our work on Structured Physical Intelligence at the MRR 2025 workshop at ICCV 2025! 🗓️ Monday, October 20 | 8:35 📷Room 308 B 🎉🤖 I'll share on how robots can perceive, reason, and adapt in the physical world. 🌍 @ICCVConference

1

2

7

Douglas Gray

@de3ug

2 months

Why attend MRR? In the age of GenAI, retrieval is a key component in grounding AI responses in reality. Now multiple fields are converging on this one subject. https://t.co/yMbYi7k4Rb

Andrew Drozdov

@mrdrozdov

3 months

More than 1100 #ICLR2026 submissions mention "retrieval" in their title or abstract 🤯

0

1

Douglas Gray

@de3ug

2 months

Why attend MRR @ #ICCV2025? Multimodal retrieval, RAG, and agentic AI—plus coffee + posters (9:50–10:35). Mon 8:30–12:30 (Room 308B). #Multimodal #RAG #Retrieval #ICCV

1

Douglas Gray

@de3ug

2 months

For full program see:

mrr-workshop.github.io

A workshop on multimodal representation and retrieval at ICCV 2025, focusing on multimodal data in applications like e-commerce, social media, and short videos.

1

0

Douglas Gray

@de3ug

2 months

At #ICCV2025? Come join us for the #Multimodal #Retrieval & #Representation (MRR) Workshop -> Mon Oct 20, 8:30–12:30 (Room 308B, Honolulu). Keynote talks from @CordeliaSchmid, @jw2yang4ai, Kristen Grauman, plus an invited talk from Roi Herzig. #AI #ComputerVision

2

3

5