Instance and promptable segmentation with Meta SAM 3.1 😍
SAM 3.1 introduces object multiplexing, which improves video segmentation performance by enabling the model to track multiple objects simultaneously in a single pass, rather than processing each object sequentially.
It is now available for quick testing in the sam3-inference repository. I am also currently working on video promptable concept segmentation (PCS), which will be added to the repository soon.
Feel free to try it out and share your feedback in the comments.
#sam3#computervision#machinelearningproject
Just wondering how long it'll take to train a 20M model on my MacBook Pro! 💻 Might be a while, but I'm curious to see the progress. Hoping for some decent 🔥 Anyone else experimented with training larger models locally? #MacBookPro#ModelTraining#machinelearningproject
Depth estimation using **MiDaS** 🔥
MiDaS is a deep learning model that infers relative depth rather than absolute metric values. It captures spatial relationships, what’s nearer or farther, using neural networks trained on diverse datasets.
👇👇
#machinelearningproject#midas
AI-generated Video!
This video showing a machine giving haircuts is AI-generated. It’s not real footage, and no such working machine exists.
hindustantimes.com/trending/us/ai…
ndtv.in/zara-hatke/man