Segment Anything Model by Meta AI: An Image Segmentation Model

This video is all about the 'Segment Anything Model' (SAM) released by Meta AI, which is an image segmentation model designed to simplify the process of segmenting images. It is a new AI model from FAIR (Facebook AI Research team) that can "cut out" any object, in any image, with a single click.
We will explore SAM by going through their GitHub repo and sample notebooks. We will also discuss some of the potential use cases of this powerful tool.
#ai #tech #coding

[tilkesh]

I was struggling to run the code in collab. After watching your video I realized I did not change using_colab = True. Thanks. I reached to your video just by youtube suggestion which I did not search. This is the power of AI. Each of our web activities changes the suggestions in Youtube. Your video and the Youtube suggestion helped me. Thanks

[CROscarAbrahamJosePadillaSolis]

Thanks for sehearing your knowledge, in google fix very well, I tried to run in local but I dont know if my gpu is not enough (6 gb Vram), when I run the model it says memory allocated is not enough, trying to reserve 1024 mib (4.6 gb reserved to pytorch), searching I found i must create a variable for the batch size, it would be great if could share how run locally or if you know some guide to adjust the use of the gpu i really apreciate it

[fortunoef]

thank you., nice video sir, . I have a project with SAM, on how to label masking images for training datasets to read animal behavior in images. thank you

[mpseiib]

Hello, I enjoyed your video! I am wondering couple of things...this algorithm has been trained on many many images and types of images are very broad but they probably didn't include many images of lung tumor. So, can we retrain it and make it more effective for lung tumor recognition? Also, could we remove the background so it only recognize the lung and the then tumor inside?

[THEGMINGLEGEND]

please cover more AI stuff .. you earned a sub

[edmald]

Greet video, but how to get the labels pixel by pixel?

[rizkamilandgamilenio]

can we train sam for our custom dataset?

[shamukshi]

i am working on "solar panel detection and counting using SAM"..it is masters level project....can you help me?....i will pay you