Assistant Professor @JHUCompSci, @HopkinsDSAI Past: RAP @TTIC_Connect | PhD @SiebelSchool Research: Pixels — Perception — Physics
Baltimore, MD
Joined June 2011
- Tweets1.7K
- Following523
- Followers3.7K
- Likes10.4K
Pinned Tweet
🧵[1/n] We often hear that breaking things is easier than building them. But can we teach a video model how things break?
Excited to share FracGen, a physics-informed and controllable framework for fracture-video generation. FracGen learns how objects stretch and tear from a single image under specified physical conditions.
Work done by my PhD students @Tung11199951 and @jiahan_zhang1 @JHUCompSci
Paper: arxiv.org/abs/2609.38152
Project: fracgen.github.io/
Anand Bhattad retweeted
New preprint:
Cross-modal steering vectors transfer zero-shot in both directions 🖼️ ↔️ 📝
From text generation to image generation and vice-versa
Exciting evidence for a shared conceptual space!
Fantastic effort led by Donghao Zeng and awesome co-authors @nsubramani23 & @TamarRottShaham
Details 🧵👇
Anand Bhattad retweeted
Come check out OpenTac at CoRL!
We will be demoing OpenTac at #CoRL2026 this year! Check out @arikslep's thread for more details.
nitter.cf/arikslep/status/211295…
We will be demoing OpenTac at #CoRL2026 this year! Check out @arikslep's thread for more details.
nitter.cf/arikslep/status/211295…
We have 17 exciting demos at #CoRL2026 this year! 🤖 From humanoids and robot foundation models to dexterous manipulation and tactile sensing—come see them in action!
Check out the lineup: corl.org/program/demos
Anand Bhattad retweeted
What if robot touch sensors could be designed for almost any shape, fabricated at scale, and transferred across platforms?
OpenTac introduces an open ecosystem for customizable 3D tactile skins, combining a scalable resistive sensor with tools that generate fabrication files from 2D outlines, developable 3D surfaces, and even text-based descriptions.
OpenTac decodes both normal force and local shear, while providing simulation, cross-sensor transfer, and adaptive data acquisition for control and learning. The authors designed and fabricated more than 30 sensors across robot embodiments, demonstrated tactile feedback and manipulation learning on four real-world robots, and achieved 19.2× faster partial readouts than full-array scanning on a 1024-taxel skin.
Title: OpenTac: An Open Ecosystem for Customizable, Scalable, and Transferable 3D Tactile Sensing
Project URL: opentac.github.io
Paper: arxiv.org/abs/2610.31661v1
What a wholesome video from @TTIC_Connect. TTIC is an amazing and unique academic place.
Like Greg in this video + my brother (who was visiting me then) said, "This feels like a startup environment for academics." It is exactly how I felt during my two years as an RAP at TTIC!
What does it take to advance fundamental #computerscience? In our new video, hear from TTIC faculty and students and step inside our research spaces, from #ML and #robotics to natural language processing. Watch on YouTube: buff.ly/khGNEC1
Anand Bhattad retweeted
What if you could give almost any robot a sense of touch?
OpenTac is an open-source toolkit that lets you design custom tactile sensor skins for robot fingers, grippers, hands and even robot feet.
Its CAD tool converts a 2D outline or 3D model into fabrication-ready flexible PCBs and 3D-printed sensor covers. The sensors detect both pressure and shear, helping robots understand when they're touching, pushing or slipping against an object.
Its adaptive readout reaches 1286 partial updates per second on a 1,024-sensor array which is quite impressive.
What would you build if your robot could feel?
Sharing our new work OpenTac; a tactile sensor + ecosystem that let’s you automatically design, customize, and deploy tactile skins on any robot surface!
Work led by @arikslep and in collaboration with a large group of robotics faculty new-and-not-new @JohnsHopkins. This also is my first preprint since I started my faculty position @JHUCompSci @HopkinsDSAI and setup SciPhy lab this Fall.
OpenTac is a piezoresistive robot skin that can accurately estimate normal and shear forces, can be adapted to sensorize any developable robot surface, and based on how sparse your contact is, adaptively readout at kilohertz-rate.
All of our software, hardware, and firmware is made available under open, permissive licenses for research or commercial use. 🧵
Anand Bhattad retweeted
For contact-rich interactions, check out our recent work on OpenTac, a new tactile-sensing ecosystem designed with the same principle in mind: representations can transfer across different robot morphologies.
nitter.cf/arikslep/status/211295…
Anand Bhattad retweeted
OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite R..
arxiv.org/abs/2610.31645
---
Newsletter morrislee1234.wixsite.com/we…
More story linkedin.com/in/morris-lee-p…
LinkedIn morris.short.gy/linkedin
#AINewsClips #AI #ML #ArtificialIntelligence #MachineLearning #ComputerVision
Anand Bhattad retweeted
Guruprasad et al., "OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite Ray Fields"
Knowing how rays work and putting that into the design allows you to do feedforward geometry estimation for any type of camera.
Anand Bhattad retweeted
Monocular 3D reconstruction assumes pinhole cameras — fisheye and panoramic images get distorted. OmniRay composites radial, spherical-harmonic, and free-form ray fields to handle any lens from a single shot
📄 arxiv.org/abs/2610.31645v1
🌐 nam1410.github.io/omniray/
Anand Bhattad retweeted
How should we approach research, when the very field we work on is going through rapid and potentially fundamental changes?
I recently gave a talk at the Stanford Vision Seminar, reflecting on my PhD journey and how AI is changing the way I think about research. I adapted part of it into a new blog post, organized around three questions:
1. What should we work on?
2. How should we work with AI?
3. What is a PhD for?
I hope some of my reflections are helpful. The talk ended with an open discussion, and I’d love to continue it here; curious how others are thinking about these questions.
Personally -- as much as there's FOMO, I also think life is very long, and we have time.
toruowo.github.io/blog/posts…
Anand Bhattad retweeted
OmniRay: Camera-Agnostic Monocular 3D Reconstruction with Composite Ray Fields
@gnam1410, Abhay Yadav, Cheng Peng, Rama Chellappa, @anand_bhattad
tl;dr: no intrinsics or camera-type input
arxiv.org/abs/2610.31645
Anand Bhattad retweeted
Hi all, I have a new result I'm excited to share. It turns out that *all existing implementations* of softmax and cross-entropy (that's JAX, PyTorch, textbooks, etc) have a subtle bug --- two bugs, actually. A simple rewrite fixes both, and eliminates "slingshots" when grokking:
Anand Bhattad retweeted
Can we improve vanilla pixel diffusion transformers without complicating the backbone?
Studying JiT, we identified two problems:
1. Limited model capacity favors low-dim targets like x₀-prediction.
2. JiT architecture exhibits a high-frequency deficit.
We investigate both. 🧵
Anand Bhattad retweeted
To further stress-test our sensors, we ran multiple experiments where we kept indenting the surface of our sensors, ranging from 7 hours to 12 hours. After the first ~500 presses or so, the sensor readings stabilize, and remain consistent over the entire test window.
Anand Bhattad retweeted
Amazing work by the Johns Hopkins team on tactile sensing and compliant control! Proud to support researchers pushing the boundaries of what’s possible with Physical AI through our Research Grant Program.
We strapped an OpenTac skin on the forearm of a @DexmateAI Vega-1 robot. We can sense external contacts and nudge the robot around by implementing a compliant controller that monitors pushes and complies with the direction of each push.
Anand Bhattad retweeted
Tactile sensing is one modality that struggles intensely with data scarcity. That's why, I think efforts to unify hardware across the field is very valuable for future research. Very exciting to see this! @arikslep
Anand Bhattad retweeted
If the barriers to entry for tactile sensing for you are:
- "It's too hard -- I need to have mechanical and electronic design expertise",
- "Aren't the sensors usually very fragile?",
- "I don't know where even to start",
- "Assembling and customizing sensors is a lot of work", or
- "It's very expensive. Who'd pay $500 for a fingertip sensor that wears out in a week of heavy use?"
then you'll want to try OpenTac. @arikslep and team worked extra hard to ensure everything 'just works'
The hardware barrier to entry is gone. Instead of manually engineering sensor arrays, developers just upload a 3D model of their robot part into OpenTac CAD.
Within three seconds, the software automatically routes a custom flexible printed circuit board and generates files for a 3D printable protective bump case, covering up to 93 percent of the target area.
Anand Bhattad retweeted
Most robotic hands are completely numb because wrapping flat tactile sensors around curved geometry is physically impossible.
@JohnsHopkins and @MIT just fixed this with OpenTac.
This open source ecosystem automatically designs custom 3D tactile skins for any robot surface, delivering normal and shear force feedback at a blazing 1286 frames per second.
Anand Bhattad retweeted
The hardware barrier to entry is gone. Instead of manually engineering sensor arrays, developers just upload a 3D model of their robot part into OpenTac CAD.
Within three seconds, the software automatically routes a custom flexible printed circuit board and generates files for a 3D printable protective bump case, covering up to 93 percent of the target area.