deploy Token FiftyOneAI $FOAI on @base Live in @bankrbot
Github : https://t.co/DJRg6UacYM……
Website : https://t.co/2Sz3IrnVvX
We created FiftyOne to supercharge your visual AI development by enabling you to visualize and label.
hey there!
We created FiftyOne to supercharge your visual AI development by enabling you to visualize and label your data, evaluate your models.
Website : https://t.co/ijXspTmffw…
Github : https://t.co/LMpoEyGL4A
We’re heading to #WACV2026 in Tucson to share how teams can maximize AI performance with better data.
Here’s what we’re bringing:
🎤 Plenary panel: Our CSO & co-founder @JasonCorso will join speakers from @Amazon, @VCU, and @Intelligence and National Security Alliance to discuss the gap between academic benchmarks and the realities of deploying CV systems in production. Join the panel on 3/8, 12:30-1:30 PM at AZ Ballroom Salon 6.
📄 Research paper: Our ML team is presenting new research on Zero-Shot Coreset Selection via Iterative Subspace Sampling — a method for automatically identifying the highest-impact samples from raw, unlabeled data so you can focus labeling efforts where they matter most.
📍 Booth #11: Stop by to connect with our team and learn how better data can help maximize your model performance.
See you there! https://t.co/LzFmT7ZQAB
Voxel51, @NVIDIA, and @nebiusai are partnering to power Physical AI Data Factory: an open reference architecture that unifies how physical AI training data is generated, augmented, and evaluated at scale. Check out the first-ever live demo at #GTC 2026, booth #1645.
Synthetic data has been the promise to solve one of the hardest problems in the physical AI space: generating the rare, high-stakes scenarios that you can’t reliably capture in the real world. But most pipelines produce outputs with subtle errors: incorrect traffic signal states, lighting inconsistencies, hallucinated scene elements. The pipelines that do generate accurate data simply haven’t existed outside of the largest research labs.
The Physical AI Data Factory Blueprint changes that. It gives development teams a way to move from raw data to model-ready training sets, closing the long-tail distribution gaps that real-world collection can't reliably cover.
@Porsche’s AV team uses embedding search in @Voxel51 to identify which scenes have the most impact on model performance. From there, it moves into the generation pipeline where @nebiusai handles the compute. The pipeline uses @nvidia Cosmos Reason or Qwen3-VL to auto-label the source footage and user can generate augmentations across weather, lighting, and traffic states. Finally, Cosmos Grader scores output across multiple axes, and only the outputs that pass this get surfaced back in FiftyOne for review.
If you’re at #GTC2026, check out the demo at booth #1645. Learn more about the Physical AI Data Factory: https://t.co/uyrMC1eP6o
@NVIDIADRIVE@nvidiaomniverse@NVIDIAAI
One of the biggest challenges for autonomous driving teams is curating a set of key moments into assets to create assets that can be trusted and reused. @Porsche researchers are tackling this problem head-on with a unified engineering workflow that connects data curation, audit, reconstruction, and scenario expansion into a single pipeline with the power of Voxel51, @nvidiaomniverse, and @NVIDIAAI Cosmos.
This structured audit process eliminates up to $350K in annual preventable spend from bad data training incidents. Neural reconstructions compress manual effort into a few hours with a 10x expansion in scenario diversity without additional fleet miles.
Read a blog post from @Porsche researcher, Tin Stribor Sohn, about how they achieved fewer wasted reconstruction cycles, faster turnaround from capture to simulation, and a regression library that grows more robust over time: https://t.co/mC39wD0DsY
Physical AI teams aren't short on 3D data, but lack a lightweight way to visualize it at full fidelity: downloading multi-GB scene files, spinning up cloud GPUs just to review a reconstruction, waiting for assets to load before you can do anything useful.
That's the problem Voxel51 and @Miris_spatial are tackling together. Miris is integrating its spatial streaming SDK into FiftyOne so teams can view full-fidelity 3D scenes alongside their annotations, model outputs, and metadata — no cloud GPU required.
Learn more and sign up for early access: https://t.co/7oGEAoUIfw
It’s clear: The next wave of visual and physical AI progress will depend less on building bigger models and more on building better data operations.
Read the full report at the link below!
https://t.co/lvt14WeHLS
100% of respondents in our new survey reported underperforming visual AI models.
Not 90%. Not “most.” Every single team.
And according to 89% of the 700 visual and physical AI professionals we surveyed, the primary reason comes down to one thing: data.
Learn more 👇
https://t.co/8wb7jM7v1v
Voxel51 is heading to #ICRA. Stop by booth 81 to see the latest FiftyOne demos, including native multimodal data support. Ingest MCAP data, play back multi-sensor scenes, search across petabyte-scale datasets, and curate data for model training, all in one platform for your robotics workflow.
Also, don't miss @DataScienceHarp 's Tech Talk at Hall C7, on 6/3 at 10 AM.
See you in Vienna, DM us to connect.
static 3D reconstruction is mostly solved. dynamic scenes, where objects move and people walk around, that's still an open problem.
the bottleneck is data: you need multiple synchronized cameras capturing the same moment from different angles with dense ground truth
Syn4D is a fully synthetic multiview dataset built for this. 8 synchronized cameras, Unreal Engine 5, per-frame depth maps, instance segmentation, camera poses, and natural language captions across offices, warehouses, and hospitals
i grouped the 8 camera views together in fiftyone with 3D point cloud reconstructions so you can flip between any camera angle, the depth and segmentation overlays, and the fused 3D scene for any sequence
check out the dataset here: https://t.co/0tmVGflYQR
btw if you're at ICRA next week hmu or come by booth or swing by booth B081 and say hi
#ICRA2026
Day 1 at #CVPR2026. The Voxel51 team was out across the workshop floor today: Sid Mehta at FGVC Workshop, Murilo Gustineli at Latinx in AI (LXAI), and Jason Corso at both Auto-Expert and CV in Agriculture.
If you missed us, find us at 📍 Booth #309 June 5–7. We'll have flash talks with speakers from @NVIDIA, @BoozAllen, Miris, and Voxel51.
Learn more about what we're doing at #CVPR2026: https://t.co/hi0JijZgZp
What is DuckyClaw?
A hardware-oriented AI Agent framework. Not just a dev board—a full deployment stack for the physical world.
When your AI Agent needs to perceive, decide, and control real devices, DuckyClaw is the stack. Built on TuyaOpen C SDK—no Node.js on MCU.
Local agent loop + Tuya Cloud
Start on device, scale with the cloud. One agent loop, unlimited possibilities.
Don't limit your agent to local—get more with Tuya Cloud.
Build your own hardware skills
Create and extend hardware skills. Connect cameras, sensors, and other peripherals with deep integration into the agentic AI loop.
testing something with @pawthereum
$100k in an endowment sitting in @mamo
each week, the @bankrbot claims yield & posts a 24h poll on X for which animal shelter gets the donation
funds are then donated via @endaomentdotorg
all tracked in a bankr app
Today, LienFi officially minted the first tokenized U.S. property tax lien on @base.
This marks an important milestone for LienFi and the broader real-world asset ecosystem.
We are proud of the work our team has done to bring one of America’s oldest yield markets onchain.