Luowei Zhou

Ex-DeepMind. Ex-Microsoft. UMich Ph.D. Ex-Intern@FAIR, MSR, Metamind. Opinions my own.

[Luowei Zhou @luowei_zhou]

Our ClipBERT just won CVPR 2021 Best Student Paper Honorable Mention! We are truly honored. Thanks to all authors! #CVPR2021 #CVPR2021BestPaperAwards

arXiv: https://t.co/PVnhPAZYoJ
Code: https://t.co/pYDGlI66xe

@jayleicn @linjiefun @zhegan4 Tamara @mohitban47 JJ @msftresearch

[Luowei Zhou @luowei_zhou]

Wow so artistic & cool! This is a game changer on real-time video generation! A big step towards real-time interactive video/world creation. Congrats team!
@XilinZhang6

[Daydream @DaydreamLiveAI]

Today’s the day🚀

We’re opening the doors to the Daydream platform - API + Playground Experience🌊

We’re the world’s first hosted StreamDiffusion platform:
1⃣ No GPU required
2⃣Sub-second latency
3⃣Advanced controls that let you experiment quickly and scale seamlessly.

This is the start of an open ecosystem for real-time video and world models. Your home for open source, real-time video AI 👇🧵

[Luowei Zhou @luowei_zhou]

Super cool! Daydream transfers your live feed into any style. Try your own prompt today!

[Jason Corso @_JasonCorso_]

🚀 The Corso Group at the University of Michigan is Recruiting PhD Students! 🚀

Are you passionate about pushing the boundaries of AI, machine learning, and computer vision? Does the lack of fundamental understanding about how data distributions lead to certain properties in the resulting model nag at you? Are you driven to improve how AI and Humans naturally collaborate in the physical world?

The Corso Group (COG) at the University of Michigan is recruiting 2-3 talented, self-motivated, and creative PhD students for the 2024-2025 academic year.

Led by Prof. Jason Corso, COG has been pioneering advances in physical AI and visual AI for the last two decades. We've contributed seminal work in areas such as machine learning foundations, video understanding (including the first paper on video captioning), human-in-the-loop computer vision, and interactive physical systems.

🔍 Current Research Areas:
- Human+AI collaboration to solve real-world challenges in the 3D world, including rural healthcare and mechanics (e.g., DARPA PTG program where I'm the PI: DARPA PTG https://t.co/U3MW52d9hp, DARPA News https://t.co/o5jb8Ozd4A).
- Foundational machine learning: understanding the relationship between data, datasets, embeddings and models; enforcing certain structure during model learning to enforce problem-relevant properties; adaptive learning through self-guided inquiry with no labeling.
- Procedural and instructional computer vision in goal-oriented physical AI.
- Bridging the sim2real gap.
…and many more exciting challenges!

🌐 We encourage applications from individuals of all backgrounds, races, genders, identities, ages, and degrees, both domestic and international. Students can be recruited through CSE, ECE, and ROB PhD programs.

🎓 Interested?
Submit your interest by completing this form by Dec. 1. https://t.co/AeR5r4vjz7

📢 Note: While I also serve as co-founder and Chief Scientist at Voxel51 https://t.co/g9AF9ngQJW (creators of the open-source FiftyOne toolkit), these PhD positions are unrelated to Voxel51. Learn more here: https://t.co/bAVBmaJOxz and https://t.co/3tVQ0l7oty

[Demis Hassabis @demishassabis]

As an AI-first company, I'm so excited for what's coming, including Gemini, the new Google DeepMind foundation model in training that will have impressive multimodal capabilities not seen in prior models. Can't wait to share more in the future! #GoogleIO

[Sundar Pichai @sundarpichai]

3/ Bard is now running on PaLM 2, and we continue to evolve it with coding upgrades, export options to @gmail @googledocs, Japanese + Korean language support (with more on the way), the ability to ask questions using images with Google Lens + more.
https://t.co/Z86q8nZsTK

[Sundar Pichai @sundarpichai]

2/ Bard seeks to combine the breadth of the world's knowledge with the power, intelligence, and creativity of our large language models. It draws on information from the web to provide fresh, high-quality responses. Today we're opening Bard up to trusted external testers.

[Luowei Zhou @luowei_zhou]

Visual Clues: https://t.co/vn3eNtgtcf
("The Economist" coverage: https://t.co/gci3DCAsjG)

GPT-3 for video temporal understanding: https://t.co/qftcqdGFts
(Code: https://t.co/NS9NEDwzaF)

Video-image-text pretraining: https://t.co/ODEfSdhAWb 2/2

[Luowei Zhou @luowei_zhou]

We are presenting three papers at #neurips2022 on vision-language foundation models. Please check out our poster sessions if you are interested! Summary: 1/2

[European Conference on Computer Vision #ECCV2026 @eccvconf]

#ECCV2022 list of paper IDs of accepted papers is NOW available!

https://t.co/bTJdeqGBYU

[Min-Hung (Steve) Chen @CMHungSteven]

Do you know there are 250+ Transformer-related papers @CVPR this year?
Check out this paper list (including papers, codes, websites, etc.) to avoid missing any!
https://t.co/kMThHeO7Gg
Feel free to share with others😀
#CVPR2022 #MachineLearning #ArtificialIntelligence

[Luowei Zhou @luowei_zhou]

We are on @TheEconomist! The "Orchestration" of Foundation Models (in our case, Microsoft Florence + OpenAI GPT-3) makes machines creative.
Blog: https://t.co/gci3DCAsjG
More examples: https://t.co/mFqFgF9Kdr
Paper/technical details on Visual Storytelling: https://t.co/nRiR7aQRMa

[Luowei Zhou @luowei_zhou]

Sunset or sunrise? Not a problem. As a step towards understanding temporal dynamics with foundation models, we propose VidIL, a few-shot video-text learner w/o training on video. We see promising results vs. Flamingo on video QA and beat sup. SOTA on Video Event Prediction (VLEP)

[Zhenhailong Wang @zhenhailongW]

Can GPT-3 understand videos? Glad to share our new work VidIL on prompting LLMs to understand videos using image descriptors (frame caption + visual token). We show strong few-shot video-to-text generation ability WITHOUT the need to train on ANY videos: https://t.co/96zmEqGe1n

[Luowei Zhou @luowei_zhou]

Don't let anything get in the way of your graduation moment! Feel surreal to reunite with friends after two whole years of the pandemic. Thanks @ProfJasonCorso for believing in me all along and the mentorship (and many many others). What a journey! And as always, go blue!!

[Jason Corso @_JasonCorso_]

Big day for COG! Congratulations to @BanerjeeShurjo (ECE) Ryan Szeto (CSE) @luowei_zhou (ROB) Kyle Byungsu Min (ECE) for persevering through the pandemic to finish their PhDs @UMRobotics @UMichECE @UMichCSE

[Michigan Robotics @UMRobotics]

Announcing the first robotics department among top-ten engineering schools:

The @UMengineering Robotics Department!

Read more: https://t.co/cPdXJzWpwo

[Luowei Zhou @luowei_zhou]

@visarga Great point! We haven't tried it yet. It would be interesting to see it on DALL-E/NUWA. Please stay tuned.

[Luowei Zhou @luowei_zhou]

So proud of the team!! Florence not only works incredibly well on a variety of image tasks (classification, detection, etc.) but also video tasks (action recognition, retrieval), often right out of the box w/o any fine-tuning. Stay tuned for more on our vision foundation models!

[AK @_akhaliq]

Florence: A New Foundation Model for Computer Vision
abs: https://t.co/sxNJitYcL0

sota results in majority of 44 representative benchmarks, ImageNet-1K zero-shot classification with top-1 accuracy of 83.74 and the top-5 accuracy of
97.18, 62.4 mAP on COCO fine tuning

[Luowei Zhou @luowei_zhou]

@Mlbot4 @ak92501 It's still under discussion. Please stay tuned!

[Luowei Zhou @luowei_zhou]

@laion_ai @ak92501 It's still under discussion. Please stay tuned!

[Luowei Zhou @luowei_zhou]

@ManlingLi_ @MSFTResearch @elgreco_winter @JiaweiHan @Shih @ChenguangZhu2 @RuochenXu1 @shuohangw Congrats Manling! It's a pleasure to work with you!

[Luowei Zhou @luowei_zhou]

Come and join us in the VALUE Challenge (for Video-And-Language Understanding Evaluation)! Winners to be announced at the CLVL workshop @ICCV_2021

More details👉https://t.co/8U3x8KljRK

[Linjie (Lindsey) Li @LINJIEFUN]

🎉Our VALUE paper has been accepted to NeurIPS 2021 Dataset and Benchmark Track.

Only 25 days left for the VALUE Challenge 2021!

Participate to win up to $22.5K prizes!

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