Segment Anything Model (SAM) - Foundational Model Deep Dive

Big thanks to UPDF who are sponsoring the video, you can get the PDF editor I was using in this tutorial over here: https://bit.ly/4f9DwGK

Image segmentation was historically hard manual work.

In today’s deep dive video, I’ll show you the methodology that Meta used to start a segmentation foundational model called Segment Anything which can achieve great zero-shot results in multiple computer vision tasks.

Also, for those asking, I'm using TLDRAW to analyze the code:
https://www.tldraw.com/

# Table of Content
- Introduction: 0:00
- Task: 1:17
- SAM Testing: 5:40
- Model Theory: 10:23
- Model Code Overview: 14:16
- Image Encoder Code: 20:53
- Prompt Encoder Code: 22:29
- Mask Decoder Code: 25:37
- Data & Engine: 37:25
- Zero-Shot Results: 39:50
- Limitation: 42:24
- Conclusion: 44:11

Here are a few interesting links to dive further into SAM:
📌 Code: https://github.com/facebookresearch/segment-anything
📌 Paper: https://arxiv.org/abs/2304.02643
📌 Great blog post on the other SAM variants: https://www.lightly.ai/post/segment-anything-model-and-friends

----
Join the newsletter for weekly AI content: https://yacinemahdid.com
Join the Discord for general discussion: https://discord.gg/QpkxRbQBpf

----
Follow Me Online Here:

Bluesky: https://bsky.app/profile/yacinemahdid.bsky.social
LinkedIn: https://www.linkedin.com/in/yacinemahdid/
___

Have a great week! 👋

Тэги:

#programming #machine_learning #data_science #optimization #feature_engineering #deep_learning #neural_networks #artificial_intelligence #data_visualization #foundational_models #segmentation #automatic_segmentation #single_point_mask_segmentation

Скачать видео