Founder, Spatial Support — 3D Commerce Powered by Machine Intelligence (Voice/Chat, Spatial Analytics)
the middle of entropy & order
Joined January 2021
- Tweets2K
- Following819
- Followers449
- Likes655
Joseph aka Joe retweeted
In the last 6 months the LocalAI team have created 17 C++ inference backends:
depth-anything.cpp (from the video) — Estimates depth from ordinary images and can produce depth maps or 3D point clouds.
Joseph aka Joe retweeted
A key thing limiting reasoning about 3D human-object interaction in images is the lack of training data. In PICO, we propose a novel way to label human-object contacts in images and build a new dataset with dense, vertex-level, contacts on the humans and objects. We then infer the 3D object shape and pose, human shape and pose, and use contacts in an optimization method to solve for a consistent 3D configuration.
If you’re not in freefall, you’re likely contacting something. Yet 3D human-object interaction (HOI) reconstruction remains underexplored. PICO (#CVPR2025) recovers humans 🏃♂️, objects 🏓, and their interactions 👉🍎- all in 3D, from just a single internet image. 1/11
Joseph aka Joe retweeted
I’d love to humbly introduce you to something we believe is going to be a game changer for both VR and VFX.
When I first tried it I couldn’t take a headset off for like 10 mins. The hair, the dress, the movements — your brain just can’t stop being mesmerized.
Tech thread 🔽
Joseph aka Joe retweeted
Paul Graham: "There are as far as I know no books specifically about founder mode."
@omniscience42: "Hold my beer." 👇
Joseph aka Joe retweeted
Founder Mode: paulgraham.com/foundermode.h…
Joseph aka Joe retweeted
I will soon publish a new library to handle generic dynamic BVH (for frustum culling and raycasting), already used by my 'three.ez/instanced-mesh' library!🥳
repo: github.com/agargaro/bvh.js
instanced-mesh: github.com/three-ez/instance…
#threejs #webgl #webgl
Joseph aka Joe retweeted
🔥🔥We're excited to introduce MeshAnything v2, our latest research in auto-regressive mesh generation! It outperforms our previous method in 3D asset complexity and quality. Check out the paper, code (MIT license), and HuggingFace demos here: buaacyw.github.io/meshanythi…
Joseph aka Joe retweeted
Multi GPU support has been added to @nerfstudioteam with the release of gsplat 1.2!
🌟gsplat🌟 (docs.gsplat.studio) now supports multi-GPU distributed training, which nearly linearly reduces the training time and memory footprint.
Now Gaussian Splatting is ready for the city-scale reconstruction!
kudos to this amazing paper: daohanlu.github.io/scaling-u…
Joseph aka Joe retweeted
One night, at Page Street, a lady came by and threw a rock through our window. Next day it was the neighbors window. Next day, another neighbors, and so on, for weeks.
I cleaned up the glass for the old lady on rent control next door. Every window was broken and boarded up.
Once, she threw a bottle at a window, in front of the police, screamed "what are you going to do, arrest me?", and then they didn't.
I tried to get a restraining order, but didn't correctly have all 3 copies of the right forms, and was told to come back with a lawyer. I tried talking to her, but she's not even coherent.
That house produced Streamline, that helps clean energy projects get grants, Exa, an AI competitor to Google, and Brev, which helps hackers train models.
Business owners say they’re tired of constant harassment and assault from mentally ill and addicted offenders. sfstandard.com/2024/08/03/sa…
Joseph aka Joe retweeted
As someone who lives in china and worked in manufacturing - people have no idea how cheap it actually is to make things. My apartment is stacked with tech that would cost at least 8x stateside.
RT @langchain: LangChain 🤝 Nvidia NIM APIs
Nvidia just launched NIM APIs for a HUGE selection of AI models
They are all easily usable w…
Joseph aka Joe retweeted
The hackathon of the future.
🥽 Calling all visionOS developers!
We're hosting Vision Hack, the first global visionOS hackathon, September 13-15!
Sign up here: visionoshackathon.com/
Details below 👇
Joseph aka Joe retweeted
.@Shutterstock just launched the first ethical generative 3D API, and the underlying model was trained entirely on licensed data to ensure fair compensation for the original creators. shutterstock.com/press/Shutt…
Joseph aka Joe retweeted
Introducing Meta Segment Anything Model 2 (SAM 2) — the first unified model for real-time, promptable object segmentation in images & videos.
SAM 2 is available today under Apache 2.0 so that anyone can use it to build their own experiences
Details ➡️ go.fb.me/p749s5
Joseph aka Joe retweeted
RE: the degree to which high fidelity, photorealistic avatars are a prerequisite to widespread adoption of spatial collab/communication-based use cases specifically - it's probably important to consider how important of a factor inter-generational preferences are vs just the tech.
Today, you have a not insignificant portion of Gen Zs/Alphas who are effectively being socialized to the idea of interacting via lower fidelity and/or unrealistic avatars thru platforms such as Roblox, Fornite, RecRoom, and even VR-native games like Gorilla Tag, etc.
I wonder if they will *need* photorealism to interact via VR, when they age into the workplace? 🤔
Joseph aka Joe retweeted
At #SIGGRAPH2024 tomorrow, be sure to attend the Khronos Fast Forward to hear the very latest updates from the Khronos working groups that are creating the standards to enable your next application or project!
All Khronos sessions can be found at: khronos.org/events/siggraph-…
