Correll Lab

Medium: Read and write stories. On Medium, anyone can share insightful perspectives, useful knowledge, and life wisdom with the world.

Any opinions on "Toward Data Science" moving off of Medium.com? Wonder how many authors have published on TDS because it was on Medium.

Deep Seek just released the multi-modal model Janus-Pro github.com/deepseek-ai/...

Dataset to Improve Performance in Open World Manipulation The transformer architecture has changed the field of Artificial Intelligence forever in a sweeping fashion by bringing forth a new…

End-to-end learning datasets should include negative examples, check out James Watson’s block-stacking dataset medium.com/correll-l...

Embodied AI For Block Stacking Tasks Embodied artificial intelligence is currently an active research frontier, foundation models like ChatGPT have demonstrated incredible…

We tested out-of-the-box VLM reasoning for robotic block stacking with ChatGPT4o (95%) a huge step up, but symbolic planning is still needed medium.com/correll-l...

Integrating Contact Force for Delicate Robot Grasping Improving end-to-end learning with salient information

Robot end-to-end learning gets better when including force medium.com/correll-l...

A review on tactile end-to-end robot policies Can learning unify the tactile experience across different sensing modalities?

More and more folks discover the power of tactile/force information in end-to-end learning manipulation medium.com/correll-l...

Building CLIP from Scratch to Identify Sports Celebrities Unsupervised Learning Approach for Identifying Sports Legends in Images.

Building CLIP from scratch to identify sport celebrities medium.com/correll-l...

Draft — Using Diffusion Models for BeliefMDPs Robotics has come so far, and there are many examples of people achieving incredible tasks with them — whether it’s self-driving cars…

Check out Aritra and Himanshu’s attempt to solve BeliefMDPs with diffusion models medium.com/correll-l...

Bringing Tactile VLMs to Local GPUs Scaling the size of the OCTOPI tactile VLM down to a more digestible size

Two approaches of how to squeeze large (T)VLMs onto a 4090 medium.com/correll-l...

Spectrograms & Vision Transformers for Time Series Forecasting (Part 2) An exploration of using multiple time series instead of one to improve financial time series forecasting.

Spectrogram and Vision Transformers a powerful tool for time-series analysis medium.com/correll-l...

Training Multimodal Robot Models with DROID In this article I’ll be going through training a robot policy step-by-step with the DROID policy learning repository with RLDS format data…

A tutorial on training multi-modal end-to-end robot policies with DROID medium.com/correll-l...

Inpainting 3D Point Clouds Using Diffusion Networks 3D UNET Backbone With Adjusted Conditioning for Diffusion Networks on 3D Object Representation Generation

Using ControlNets to condition diffusion to in-paint 3D point clouds for manipulation medium.com/correll-l...

Diffusion Policy: Applying Diffusion Model to Policy Learning Explaining diffusion policy, a powerful imitation learning method and the building block for robotics foundation models

Diffusion models are emerging as an universal frontend in robotic foundation models. Find out how they work in Yutong Zhang’s summary of the original robotic diffusion paper medium.com/correll-l...

Add to that to using blogging and youtube to create accessible versions of their work that can be consumed quickly. (And focussing on a low-ceiling summary will also help during the paper creation process - if the paper cannot be summarized in one sentence why should anyone read it?)

A review on tactile end-to-end robot policies Can learning unify the tactile experience across different sensing modalities?

More and more folks discover the power of tactile/force information in end-to-end learning manipulation medium.com/correll-l...

Dataset to Improve Performance in Open World Manipulation The transformer architecture has changed the field of Artificial Intelligence forever in a sweeping fashion by bringing forth a new…

End-to-end learning datasets should include negative examples, check out James Watson’s block-stacking dataset medium.com/correll-l...

YouTube video by Unitree Robotics Unitree B2-W Talent Awakening! 🥳

Just WOW: youtu.be/X2UxtKLZnNo?...

Integrating Contact Force for Delicate Robot Grasping Improving end-to-end learning with salient information

Robot end-to-end learning gets better when including force medium.com/correll-l...

Still hoping humanoids will solve the greenfield vs brownfield problem that robots suffer also in this industry

😂

@wxie.bsky.social

Open-source implementation of robot policy with Flow Matching 🔥
github.com/HRI-EU/flow...

Great work from Fan Zhang and Michael Gienger from Honda Research Institute EU!

Great direction!

A review on tactile end-to-end robot policies Can learning unify the tactile experience across different sensing modalities?

i read five papers on tactile robot policies and wrote about them: medium.com/correll-lab/...

the first step of me writing my prelim and i hope to have some more interesting synthesis soon. this is in no way a comprehensive review and i welcome recommendations of papers to read!

Contextual Reasoning Meets Structured Planning Modeling Problem Statements with LLMs

Check out Praneeth Nemani’s work on using VLMs to provide scene understanding to symbolic planning medium.com/correll-l...

Paper Review: Set Transformers A Framework for Attention-based Permutation-Invariant Neural Networks

Neural networks do historically bad on sets, check out Himanshu Gupta’s review on set transformers medium.com/correll-l...

Toward Humanoids Ongoing pre-publication research on humanoids and related technologies including paper reviews and tutorials.

Follow and engage with our research on humanoid robots on medium.com/correll-lab

Embodied AI For Block Stacking Tasks Embodied artificial intelligence is currently an active research frontier, foundation models like ChatGPT have demonstrated incredible…

We tested out-of-the-box VLM reasoning for robotic block stacking with ChatGPT4o (95%) a huge step up, but symbolic planning is still needed medium.com/correll-l...

Embodied AI For Block Stacking Tasks Embodied artificial intelligence is currently an active research frontier, foundation models like ChatGPT have demonstrated incredible…

Cool! We just released a benchmark dataset on medium.com/correll-lab/...

Grasping With Common Sense using VLMs and LLMs How to leverage large language models for robotic grasping and code generation

Check out how ChatGPT can provide common sense knowledge for better grasping medium.com/towards-d...