Here for preprints.
Joined June 2018
- Tweets289
- Following1.1K
- Followers227
- Likes2.5K
Chaitanya (Chay) Ryali retweeted
We’re releasing SAM 3.1: a drop-in update to SAM 3 that introduces object multiplexing to significantly improve video processing efficiency without sacrificing accuracy.
We’re sharing this update with the community to help make high-performance applications feasible on smaller, more accessible hardware.
🔗 Model Checkpoint: go.meta.me/8dd321
🔗 Codebase: go.meta.me/b0a9fb
Chaitanya (Chay) Ryali retweeted
Excited to share SA-FARI which will be presented as an oral at CVPR 26!
conservationxlabs.com/sa-far…
My team at Meta collaborated with ConservationX Labs to create the largest open video dataset for wildlife detection -- with @Surisdi @wrong_whp @YuanTingHu1
Chaitanya (Chay) Ryali retweeted
New on @instagram Edits: AI-powered video effects, enabled by our new SAM3 model, make it easier to blur an object, tag an outfit, outline, and more.
about.fb.com/news/2025/04/in…
Chaitanya (Chay) Ryali retweeted
🔉 Introducing SAM Audio, the first unified model that isolates any sound from complex audio mixtures using text, visual, or span prompts.
We’re sharing SAM Audio with the community, along with a perception encoder model, benchmarks and research papers, to empower others to explore new forms of expression and build applications that were previously out of reach.
🔗 Learn more: go.meta.me/568e5d
Chaitanya (Chay) Ryali retweeted
Found the perfect sport to stress test Meta's SAM3 person segmentation.
Dense crowds, extreme motion, zero structure.
This is as tough as it gets, and SAM3 nailed it.
Chaitanya (Chay) Ryali retweeted
Tested Meta's SAM 3 on some low quality dashcam footage and expected the segmentation to fall apart, but it still picked up every vehicle and even spotted people on the roadside that I hadn't noticed at all.
Chaitanya (Chay) Ryali retweeted
typing "player with a red shirt" was all it took to train this computer vision model
just a few minutes to train the model, export to python, and create this video
i'll keep testing roboflow rapid and will report back with the results
in the meantime, enjoy the magic of Zlatan
data labeling is dead. long live distillation.
from data to object detection endpoint in 90 seconds.
link: rapid.roboflow.com
Chaitanya (Chay) Ryali retweeted
"laundry on the bed", "left light", "right couch" - are not "basic attributes", they are relationships. Easy way to see this: if you masked out other objects, e.g. the bed or one of the couches or lights, the referring phrase becomes incorrect for the target.
On the subset that does look like PCS, it can be even better than GT as shown.
Not sure what's not to buy. "Ref"COCO is a referring expression benchmark. A different task. Would you train on COCO and and deploy on a RefCOCO like task?
Chaitanya (Chay) Ryali retweeted
🚀 Video Segmentation and Object Tracking with SAM 3!
Learn how to segment and track objects in any video using text and point prompts with Meta’s powerful SAM 3 (Segment Anything Model 3)! Whether you're removing unwanted objects or adding new ones, this tutorial walks you through everything from start to finish.
✅ What You’ll Learn:
How to use text prompts for object segmentation
Use point-based prompts to add or remove objects
Easily track any object across multiple video frames
Real-world examples using SamGeo
📌 Useful Resources:
🔗 GitHub Repository (SamGeo): github.com/opengeos/segment-…
🔗 Notebook Example: samgeo.gishub.org/examples/s…
🔗 Meta SAM 3 Overview: ai.meta.com/sam3
📺 Check out the full video tutorial at youtube.com/@giswqs/videos
#SAM3 #GeoAI #Geospatial #OpenSource #Python #DataScience
Chaitanya (Chay) Ryali retweeted
🌍 Unlock powerful GeoAI workflows with SAM 3!
In this step-by-step tutorial, I demonstrate how to segment remote sensing imagery using text prompts and bounding boxes, powered by Meta’s SAM 3 (Segment Anything Model 3). You’ll learn how to run image segmentation on satellite and aerial imagery, extract objects of interest, and export the results to geospatial formats like GeoTIFF for further GIS or Python analysis.
🔗 GitHub Repository (SamGeo): github.com/opengeos/segment-…
🔗 Notebook Example: samgeo.gishub.org/examples/s…
👉 Check out the full video tutorial at youtube.com/@giswqs/videos
#SAM3 #GeoAI #Geospatial #OpenSource #Python #DataScience
SAM 3 is not a referring expression segmentation model - this is by design.
SAM 3 solves Promptable Concept Segmentation (PCS): segmenting objects using simple noun phrases (general categories + basic attributes) and optional exemplars. It's also a robust, composable primitive that works seamlessly with MLLMs. For example, SAM 3 + Gemini 2.5 Pro achieves zero-shot SOTA on RefCOCO+/RefCOCOg referring expression benchmarks.
Wondering if this wasn't sufficiently clear from the paper (our bad if so). But also wondering why only referring expression is benchmarked and not any of the benchmarks we reported on—the ones SAM 3 is actually designed for?
Chaitanya (Chay) Ryali retweeted
🚀 The SAMGeo Python package will soon support image segmentation with SAM3, bringing even more power and flexibility to remote sensing workflows. This integration will make high-quality segmentation easier, faster, and more accessible for everyone.
🔗 GitHub PR: github.com/opengeos/segment-…
Here’s a quick demo showing building segmentation using just a simple "building" text prompt. The results are incredibly promising! Stay tuned for updates.
#GeoAI #SAM3 #OpenSource #Python
Chaitanya (Chay) Ryali retweeted
Bringing state-of-the-art SAM models to products at @AIatMeta scale has been so fun! A favorite use case we worked on was using SAM 3 & SAM 3D for the “View in Room” feature on Facebook Marketplace—turn any furniture item in a listing to 3D and view it in AR before purchasing
Chaitanya (Chay) Ryali retweeted
As a part of the SAM 3 release, we’ve partnered with @roboflow to accelerate applying SAM 3 for real-world use cases. This includes automating visual labeling tasks, fine-tuning for novel use cases and deploying an API. Highlights of what you can do with SAM 3 on Roboflow🧵
Chaitanya (Chay) Ryali retweeted
SAM 3’s ability to precisely detect and track objects is helping @ConservationX measure the survival of animal species around the world and prevent their extinction.
🔗 Learn more about the work: ai.meta.com/blog/segment-any…
Chaitanya (Chay) Ryali retweeted
SAM 3 is the #1 trending repo on @huggingface! 🤩Thanks @ClementDelangue for personally informing us!
Chaitanya (Chay) Ryali retweeted
Meet SAM 3, a unified model that enables detection, segmentation, and tracking of objects across images and videos. SAM 3 introduces some of our most highly requested features like text and exemplar prompts to segment all objects of a target category.
Learnings from SAM 3 will help power new features in Instagram Edits and Vibes, bringing advanced segmentation capabilities directly to creators.
🔗 Learn more: go.meta.me/591040
Chaitanya (Chay) Ryali retweeted
Today we are releasing & open-sourcing Segment Anything 3 (SAM 3).
It is a state-of-the-art model for image & video segmentation, and builds upon the work of SAM & SAM 2.
SAM3 will also power features in Edits, Meta AI, & Facebook Marketplace soon.
aidemos.meta.com/segment-any…
Chaitanya (Chay) Ryali retweeted
🌟Thrilled to share that SAM 2 was awarded a Best Paper Honourable Mention Award at #ICLR2025, one of 6 papers recognized out of 11000+ submissions!
👏This project was the result of amazing work by an exceptional team at @AIatMeta FAIR: @vgabeur ,
@YuanTingHu1,@RonghangHu, @wrong_whp, Tengyu Ma, @HaithamKhedr,@raedle, Chloe Rolland, Laura Gustafson, @junting9, @KalyanVasudev, @alcinos26,
@chaoyuaw, @inkynumbers, Piotr Dollar, @cfeichtenhofer.
🚀It was inspiring to hear from researchers at ICLR applying SAM to a diverse array of real-world challenges—from medicine and semiconductor manufacturing to robotic navigation, enterprise data annotation platforms, and video editing. If you're using SAM for an innovative application, please reach out, we'd love to hear about it!
