Explaining the Segment Anything Model - Network architecture, Dataset, Training

In this video, I dive deep into the technical details and architecture behind the Segment Anything Model, also known as SAM. SAM is the world's first foundation model on image segmentation and is an amazing tool that can segment any image provided to it at multiple nested levels of granularity at interactive latency.

#deeplearning #computervision #machinelearning

To support the channel and access the Word documents/slides used in this video,  consider JOINING the channel on Youtube or Patreon. Members get access to scripts, slides, animations, and illustrations for most of the videos on my channel!
Join and support the channel - https://www.seevid.ir/c/avb_fj/join
Patreon - Patreon: NeuralBreakdownwithAVB

Project page: https://segment-anything.com/
Give the paper a read: https://arxiv.org/pdf/2304.02643.pdf

0:00 - Intro
1:29 - Architecture
4:50 - Interactive Training
6:30 - Dataset
7:27 - Model Architecture
12:30 - Outro

Other papers cited:
Focal Loss for Dense Object Detection: https://arxiv.org/pdf/1708.02002.pdf
CLIP: https://arxiv.org/pdf/2103.00020.pdf
Masked Autoencoders Are Scalable Vision Learners: https://arxiv.org/pdf/2111.06377.pdf

Songs:
Sunny Days - Anno Domini Beats
Wellington Coffee Shop - Dyalla
No 3 Morning Folk Song - Esther Abrami