Our team member (Zeqiang Lai) has reproduced #DragGAN and integrated it into InternGPT! Have a try! 🏄‍♂️ Demo: github.com/OpenGVLab/Inte Code: github.com/Zeqiang-Lai/Dr

1:32

4.4K views

141

Show this thread

After 15-month reviewing, our extension of UniFormer has been accepted by TPAMI🎉. In the major revision, we add our earlier exploration for building lightweight models. The simple models run so fast on CPU/GPU. Fast Demo: huggingface.co/spaces/Andy162 Project: github.com/Sense-X/UniFor

🌟 Excited about the releases of the #ChatGPT App and #Zelda game? 🚀 Check out the power of our multimodal LLaMA-#Adapter, with a performance that echoes the potential of the visual #GPT4. 💥 Stay tuned for the upcoming V2 demo, multimodal Arena, checkpoints, and much more!

OpenGVLab

@opengvlab

May 19

InternGPT can now generate images based on audio input by incorporating

@Meta

's ImageBind into our pipeline. learn more: github.com/OpenGVLab/Inte

0:57

39 views

Quote Tweet

OpenGVLab

@opengvlab

May 10

Very excited to announce our work, InternChat. We can now interact with #ChatGPT using the cursor, bringing human interaction to a whole new level. Demo built with @Gradio No OpenAI API key needed for a limited time! For the demo and more samples: github.com/OpenGVLab/Inte

Show this thread

OpenGVLab

@opengvlab

May 15

Exciting progress for FocalNet! Wonder if the new method can be applied to other players on the chart too.

@jw2yang4ai

Our InternImage at 65.4mAP is feeling the heat🔥

Quote Tweet

Jianwei Yang

@jw2yang4ai

Apr 26

Our FocalNet is shining again! Combing <700M focalnet-huge and stable-DINO, we achieved 64.8 mAP with only ImageNet-22K and Object365, without any test time augmentation! It beats EVA and only lags behind 3B InternImage, gives you the strongest reproducible object detector! twitter.com/jw2yang4ai/sta…

Show this thread

OpenGVLab

@opengvlab

May 12

Chatbot Arena meets multi-modality! Multi-Modality Arena allows you to benchmark #LLMs side-by-side while providing images as inputs. Which model would you like to see being supported? Demo at vlarena.opengvlab.com built with

@Gradio

Code: github.com/OpenGVLab/Mult

Quote Tweet

lmsys.org

@lmsysorg

May 10

Announcing the Week 2 update for the Chatbot Arena leaderboard! We've added some new models that are showcasing strong performance. Currently, @OpenAI's GPT-4 and @AnthropicAI's Claude lead the pack, with open-source models in hot pursuit. More findings: lmsys.org/blog/2023-05-1

Show this thread

OpenGVLab

@opengvlab

May 12

Demo now live! Our latest VideoChat connects video foundation models with #LLMs via a learnable neural interface in an end-to-end manner, exhibiting video deep video understanding such as "Why is this video funny?"

@Gradio

Demo & more: github.com/OpenGVLab/Ask-

1:52

42 views

Quote Tweet

@_akhaliq

May 11

VideoChat: Chat-Centric Video Understanding abs: arxiv.org/abs/2305.06355 paper page: huggingface.co/papers/2305.06 github: github.com/OpenGVLab/Ask-

OpenGVLab

@opengvlab

May 10

A quick view of our featured samples. Click the chimney and turn it into Eiffel Tower? No problem. More features to be discovered on our GitHub Page! github.com/OpenGVLab/Inte

GIF

Very excited to announce our work, InternChat. We can now interact with #ChatGPT using the cursor, bringing human interaction to a whole new level. Demo built with

@Gradio

No OpenAI API key needed for a limited time! For the demo and more samples: github.com/OpenGVLab/Inte

Quote Tweet

@_akhaliq

May 10

InternChat: Solving Vision-Centric Tasks by Interacting with Chatbots Beyond Language abs: arxiv.org/abs/2305.05662 paper page: huggingface.co/papers/2305.05 github: github.com/OpenGVLab/Inte

VideoMAE V2: We scale the already successful VideoMAE to 1 billion parameters and bringing the dataset to million-scale, resulting in new #SOTA on Something-Something and Kinetics, overcoming VRAM consumption, overfitting, and many more hurdles. github.com/OpenGVLab/Vide

OpenGVLab

@opengvlab

May 8

A great recap of recent #LLM progress. We will look into generating the computational performance benchmarks of our LLaMA-Adapter V2 like you mentioned in the blog, thanks

Quote Tweet

It was a great month for open source: So many LLMs came out that it's become quite overwhelming to keep track of it all. So, in this month's Ahead of AI issue, I am sharing resources and research insights on the latest open-source LLMs & datasets! magazine.sebastianraschka.com/p/ahead-of-ai-

OpenGVLab

@opengvlab

May 5

HumanBench is a foundation model centered around, you guessed it, humans! It can generalize across tasks such as ReID, pose, pedestrian detection, and many more, obtaining SOTA on 17 relevant datasets. project led by Prof.

@ouyang_wanli

code: github.com/OpenGVLab/Huma

0:36

24 views

OpenGVLab

@opengvlab

Apr 29

LLaMA-Adapter V2: This update allows the 65B model to surpass #ChatGPT on some questions in terms of response quality, while also outperforming #Vicuna Congratulations to our team! Code at github.com/ZrrSkywalker/L Follow us for more🫰

Quote Tweet

Pan Lu

@lupantech

Apr 28

65B LLaMA-Adapter-V2 code & checkpoint are NOW ready at github.com/ZrrSkywalker/L!

Big update enhancing multimodality & chatbot.

LLaMA-Adapter-V2 surpasses #ChatGPT in response quality (102%:100%) & beats #Vicuna in win-tie-lost (50:14).

Thanks to Peng Gao & @opengvlab! 2/2

Show this thread

2:11

13.8K views

You thought image-text tools could not get better? VIDEO-TEXT is here 🤯 Ask-Anything is a simple yet interesting tool for chatting about video with ChatGPT, MiniGPT4 and StableLM.

1:07

1K views

Ask-Anything, tool for chatting about video with chatGPT, miniGPT4 and StableLM github: github.com/OpenGVLab/Ask-

@Gradio

demo: http://106.14.223.212:7860/

1:09

35.2K views

167

608

OpenGVLab

@opengvlab

Apr 23

UniHCP is a human centric foundation model that can surpass expert models on many metrics! github.com/OpenGVLab/Huma

Replying to

and

Thanks! Yeap, here I also attach some interesting videos projects, which might bring some inspiration: Socratic Models: arxiv.org/abs/2204.00598 Ask-Anything: github.com/OpenGVLab/Ask- LaViLa: github.com/facebookresear Vid2Seq: ai.googleblog.com/2023/03/vid2se

OpenGVLab

@opengvlab

Apr 21

chatGPT + video! 👉 Ask-Anything We have combined #ChatGPT with our video understanding models, come check it out! github.com/OpenGVLab/Ask-

1:09

61 views

OpenGVLab

@opengvlab

Apr 11

Check out our new End-to-end Autonomous Driving Challenge, featuring 4 new tracks and a total of $100K award! 🔥 Baseline results will be based on the general vision model, InternImage (github.com/OpenGVLab/Inte). Learn more: opendrivelab.com/AD23Challenge. #CVPR2023 #SelfDrivingCars

OpenGVLab

@opengvlab

Mar 1

🏆New record in COCO object detection, and SOTA in 18 vision tasks, using only ONE model! Code: github.com/OpenGVLab/Inte Model, code, TensorRT deployment, and inference API coming soon🔥