One connection to the Meta Model API and Muse handles the rest: SAM 3.1 cuts out the dresser, Muse Image fills the top. Just describe what you want. @nikhilaravi
Give @Muse access to the SAM 3.1 API! You can connect Muse to your Meta Model API account just like any other custom connector — paste the API key once into the secure vault and Muse never sees it. Once connected, Muse can call everything on the Meta Model API platform: SAM 3.1, Muse Spark, Muse Image. Any time you want to remove a background or isolate an object, just ask Muse in chat. Here I wanted to cut this dresser out of a photo. Muse called SAM 3.1 to segment it, and then I asked it to use Muse Image to complete the missing regions on the wooden top where the objects had been.
Another banger use case with Muse Spark + SAM 3.1 via Meta Model API by @nikhilaravi.
Targeted object selection, powered by video object segmentation and tracking, unlocks a whole family of use cases. Anything in images and videos becomes addressable.
React-to-Anything: when watching a video, sometimes I just want to heart something in it rather than the whole video! You can build this with Muse Spark and SAM 3.1 - both on Meta Model API! Muse Spark captions what's in the scene → SAM 3.1 takes each noun phrase and predicts instance masks with a stable id per frame. React to an object and the reaction sticks to it, moving with it for the entire video. Have fun combining models across Meta's API platform!
Roman Rädle retweeted
This was a deliberate design choice to make SAM 3.1 easy to integrate within existing workflows! Great to see it is useful!
🧵 Quick cutouts with SAM 3.1 on the Meta Model API. Four scooters on a street, two-word prompt: "red scooter". Video created by Meta Muse, connected to the SAM 3.1 API.
Runs on the Meta Model API. The parser is open source, zero dependencies. Now it's your turn to cut out things.
dev.meta.ai/
dev.meta.ai/docs/sam/client-…
github.com/meta-models/meta-…
This is the composability story in one demo. Muse Spark captions, SAM 3.1 masks, Muse Image inpaints. Three models on the Meta Model API. Right-click to remove any object is such a good interface.
Right-Click Anything: a little demo combining Muse Spark, SAM 3.1 and Muse Image -- now all on @Meta Model API! Muse Spark captions the scene and generates noun-phrases. These are used to prompt SAM 3.1 which predicts instance masks for each object. Then, right-click any object mask as you would with text on any web page! “Look Up” hits Muse Spark web search for real product listings, “Copy” gives you the alpha-matted PNG, “Remove” paints the SAM 3.1 mask into the image and asks Muse Image to inpaint it away. SAM 3.1 also supports video so many more possibilities there too! Have fun combining models across Meta's API platform!
🚀 SAM 3.1 is now live on the Meta Model API. Detect, segment and track anything in images and video with a noun phrase prompt. One call, same API surface, billing and auth as the other Muse models. $2.50/1000 images, $0.20/1000 frames. Open weights still on Hugging Face!🧵 (1/2)
Meta’s Segment Anything Model (SAM) 3.1 is now available on Meta Model API, giving developers a fast and lightweight model for detection, segmentation and tracking in a single call on inference tuned for SAM 3.1's architecture.
Use a short phrase to find objects in images and video. One API call returns detections, pixel-precise segmentation masks, and identity-preserving video tracks.
Learn more and start building: bit.ly/4rDArq9
🧵 (2/2) Get started here:
SAM docs: dev.meta.ai/docs/sam/overvie…
Meta Model API: dev.meta.ai/
Roman Rädle retweeted
New tutorial | Text-prompt segmentation with @AIatMeta SAM3 ✨
Learn how to segment objects in images and videos using single or multiple text prompts with SAM3.
Watch here ➡️ bit.ly/48UI16s
#SAM3 #SegmentAnything #Ultralytics
Roman Rädle 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.
Roman Rädle retweeted
Today's update on the upcoming R interface to @Meta's SAM3:
We'll have a Shiny gadget that allows you to interactively explore and segment imagery.
Shown here in Positron: finding red cars in a parking lot at TCU, which are returned to your R session as an sf object.
Roman Rädle retweeted
i trained a computer vision model to detect football players, the ball, net, and goals scored
using roboflow rapid, SAM3, and a python script
this took me 2 hours and could be taken much further to create a sports analytics + tracking app
and now let's enjoy Pirlo's panenka penalty versus england at euro 2012
i trained this computer vision model just by asking for "tennis players", using roboflow rapid and SAM3
the model picks out Alcaraz and Sinner, while correctly avoiding the ball boys and fans
then I exported the model to python and wrote a script to filter the data, track movement history, and annotate the final video
this was a quick experiment but I could expand the project to track player speeds, distance covered, shot count, etc.
could use this for a sports analytics app that tracks live matches and calculates custom metrics
Roman Rädle retweeted
data labeling is dead. long live distillation.
from data to object detection endpoint in 90 seconds.
link: rapid.roboflow.com
Roman Rädle retweeted
🏀 Object Tracking Made Easy with SAM 3!
With just a simple text prompt like “player”, you can now segment and track objects throughout an entire video using SAM 3.
Easily customize object names, isolate specific players, or remove unwanted objects.
GitHub: github.com/opengeos/segment-…
Notebook Example: samgeo.gishub.org/examples/s…
#SAM3 #SamGeo #DataScience #OpenSource #ComputerVision
Roman Rädle retweeted
🚀 Why SAM‑3 Is a Game‑Changer in Computer Vision
Everyone’s talking about Meta’s Segment Anything Model (SAM) - but few realize how different SAM‑1, SAM‑2, and SAM‑3 truly are.
✨ SAM‑1: The breakthrough - point at anything in an image and instantly cut it out.
🎥 SAM‑2: Took it further - added video support, stable masks, and object tracking across frames.
📝 SAM‑3: Changed the game - text prompts + unified detection, segmentation, and tracking in one blazing‑fast model.
From interactive segmentation to real‑time video editing, robotics, and dataset labeling - SAM‑3 is the future of computer vision, arriving faster than anyone expected.
👉 Dive deeper into how SAM‑3 works and why it matters in our latest blog:
#SAM3 #MetaAI #ComputerVision #DeepLearning #AI #OpenCV
Roman Rädle retweeted
🚀 Why SAM‑3 Is a Game‑Changer in Computer Vision
Everyone’s talking about Meta’s Segment Anything Model (SAM) - but few realize how different SAM‑1, SAM‑2, and SAM‑3 truly are.
✨ SAM‑1: The breakthrough - point at anything in an image and instantly cut it out.
🎥 SAM‑2: Took it further - added video support, stable masks, and object tracking across frames.
📝 SAM‑3: Changed the game - text prompts + unified detection, segmentation, and tracking in one blazing‑fast model.
From interactive segmentation to real‑time video editing, robotics, and dataset labeling - SAM‑3 is the future of computer vision, arriving faster than anyone expected.
👉 Dive deeper into how SAM‑3 works and why it matters in our latest blog:
learnopencv.com/sam-3-whats-…
#SAM3 #MetaAI #ComputerVision #DeepLearning #AI #OpenCV
