Shunlin Lu

First-year PhD student at CUHK(SZ), intern at IDEA now

[Shunlin Lu @ShunlinL]

A unified learning paradigm is crucial for robotis. Check out our recent understanding of WBC.

[Weishuai Zeng @weishuaizeng]

We are excited to re-introduce our Behavior Foundation Model for Humanoid Robots, built upon a unified perspective of diverse WBC tasks, as a promising step toward a foundation model for general humanoid control.
🔗Website: https://t.co/AMDocTClqD
📜Paper: https://t.co/ema9egIJcE

[AK @_akhaliq]

Go to Zero

Towards Zero-shot Motion Generation with Million-scale Data

[Ke Fan @frankefanrank]

🚀 Our #ICCV25 paper "GO TO ZERO" is out - the first systematic exploration of zero-shot text-to-motion generation!
✅ 2M+ high-quality 3D motions
✅ New benchmark with 126 real-world prompts
✅ 7B motion foundation model
We're pushing motion generation into the zero-shot era!

[Shunlin Lu @ShunlinL]

MotionStreamer offers multi-round generation, long-term generation, and dynamic motion composition, which delivers a brand-new experience for animation users. Explore more at: https://t.co/bI9bjGFTPK and https://t.co/O3xoBs8i11

[AK @_akhaliq]

MotionStreamer announced on Hugging Face

Streaming Motion Generation via Diffusion-based Autoregressive Model in Causal Latent Space

[Ling-Hao CHEN (on job market) @Evan_THU]

🌟Introducing MotionCLR v1-preview demo: https://t.co/U0cVcWnqf4, which supports motion generation and several interactive motion editing downstream tasks.
🤝MotionCLR is supported by the latest Gradio-5 @Gradio , and thanks to the support of @huggingface!
👇Here is the demo:

[Shunlin Lu @ShunlinL]

A simple but effective motion editing way through attention mechanisms.

[Ling-Hao CHEN (on job market) @Evan_THU]

💃 Introducing #MotionCLR: MotionCLR is a motion generation model, supporting versatile interactive training-free motion editing methods.
📖Paper: https://t.co/KwZO32BWtq
🏠Page: https://t.co/WznVQdrnTA
🖊️Blogpost: https://t.co/tsm9IQf47B
💻Code: https://t.co/hpcX8PY7kT

[Ling-Hao CHEN (on job market) @Evan_THU]

[Update🚀] Today, we announce the #HuggingFace demo of 💃#MotionLLM. Thanks to the help and invitation from official @huggingface and @Gradio, @ShunlinL and I deployed our demo on HuggingFace running with A100 GPU🔥. Now, enjoy it!👇

Demo link: https://t.co/4SwWEsmqvR

[Shunlin Lu @ShunlinL]

The initial attempt of human behavior in VLM, hope to expand it into an all-knowing and all-powerful VLA model in the future. Perception module provides spatial information, the understanding module is for reasoning and planning, and the action model is for guiding the controls.

[Ling-Hao CHEN (on job market) @Evan_THU]

🔥Today, we announce the #MotionLLM💃, an LLM-based motion/video model for human behavior understanding!

📖Paper: https://t.co/LywqS257hA
🏠Project page: https://t.co/5T9OyQoXvT
🤗Demo: https://t.co/1ptbXb3Q9E
✍️Blogpost: https://t.co/O3V9fItlmt
💻Code: https://t.co/LuKnLGaZ8l

[Shunlin Lu @ShunlinL]

Amazing work. Proud of the whole team of IDEA CVR.

[AK @_akhaliq]

Grounding DINO 1.5

Advance the "Edge" of Open-Set Object Detection

This paper introduces Grounding DINO 1.5, a suite of advanced open-set object detection models developed by IDEA Research, which aims to advance the "Edge" of open-set object detection. The suite encompasses

[Unitree @UnitreeRobotics]

Unitree Introducing | Unitree G1 Humanoid Agent | AI Avatar
Price from $16K 🤩
Unlock unlimited sports potential(Extra large joint movement angle, 23~34 joints)
Force control of dexterous hands, manipulation of all things
Imitation & reinforcement learning driven
#Unitree #AI

[Shunlin Lu @ShunlinL]

The TMA can be considered as an alternative clip in the motion field. It is also useful for motion retrieval.

[Ling-Hao CHEN (on job market) @Evan_THU]

🎉 Announcing #OpenTMA

Today we released the codebase to answer "How to train TMA?" and "How to use TMA?".

The codes are available at: https://t.co/nH1wwkYiiw！

[Ling-Hao CHEN (on job market) @Evan_THU]

🎉 Announcing #OpenTMA

Today we released the codebase to answer "How to train TMA?" and "How to use TMA?".

The codes are available at: https://t.co/nH1wwkYiiw！

[Ling-Hao CHEN (on job market) @Evan_THU]

Happy to find that 🍅**TOMATO** representation is used as a new standard for motion representation!
The 🍅TOMATO representation is ablated from our extensive ablations.

[Sirui Xu @xu_sirui]

Introducing 🔥InterDreamer: Zero-Shot Text to 3D Dynamic Human-Object Interaction🔥

📰: https://t.co/WaVFTA1b6b
🔗: https://t.co/XwMTxfbajx

‼️Generate more complex HOIs with higher quality and better alignment to text by synergizing external knowledge.

🧵[1/5]

[Shunlin Lu @ShunlinL]

Amazing!!

[Hongyang Li (弘洋) @ARChaser_]

We are happy to introduce our new work 「TAPTR: Tracking Any Point with Transformers as Detection」. TAPTR is a simple and strong baseline with sota performance on many datasets while maintaining the advantage of speed.
ProjectPage: https://t.co/o3ha7EWkCV (demos and paper links)

[Shunlin Lu @ShunlinL]

Impressive T-Rex from Qing Jiang!

[Qing Jiang @mountch1cken]

🌟 Unveiling T-Rex! 🦖 An interactive object counting model that's as mighty as its namesake! With T-Rex, you can count ANY objects in images. Our online demo is LIVE at https://t.co/QABWtbtt53 . Github: https://t.co/t2uz7bVRZX

[Ling-Hao CHEN (on job market) @Evan_THU]

🔥🔥🔥Check out our #HumanTOAMTO🍅. We planted a TOMATO🍅 months ago and it is ripe now!
In #HumanTOMATO, we present the first attempt to generate whole-body motions with text description with the #MotionX dataset.
Now, have a taste of our TOMATO!⬇️
Video: https://t.co/WABZnBXSYO