Segment Anything Model

Segment Anything Model or SAM for short is one of the coolest computer vision models that we saw lately.
​
Here's how it works.
​
The model can be split into 2 parts: inputs encoder and mask decoder.
​
The inputs of the model are:
- An image.
- A prompt.
​
The outputs of the model are 3 valid masks. They represent:
​
- The whole of an object.
- A part of an object.
- A subpart of an object.
​
There are 2 types of prompts: Sparse and dense.
​
Sparse prompts can be:
​
- Text.
- Point coordinates.
- Bounding box coordinates.
​
There is only one type of dense prompts: a mask (not to be confused with the output masks!).
​
Now, let's go through how inputs and outputs are mapped.
​
The image goes through an encoder, which is a vision transformer (ViT).
​
The prompt goes through an encoder as well, but there are 2 different encoders depending on the type of the prompt:
​
- If the prompt is sparse, then the text encoder from CLIP is used to encode the information.
- If the prompt is dense (a mask), then a CNN is used to encode the information.
​
The mask decoder efficiently maps the image embedding, prompt embeddings, and an output token to a mask.
​
For more details about the model, check out Facebook's blog. Also, you can check the code on github.

​

Generative AI Courses by Google

Google just released a new learning path for Generative Machine Learning. It looks awesome!
​
Here are the Courses found in this learning path:
​
- Intro to Generative AI.
- Intro to Large Language Models.
- Intro to Responsible AI.
- Intro to Image Generation.
- Encoder-Decoder.
- Attention Mechanism.
- Transformers and BERT Models.
- Create Image Captioning Models.
- Intro to Gen AI Studio.

Check out the courses here.

Cool AI Tools (Affiliates)

​Chatbase: An AI chatbot builder that lets you build, train, and embed smart chatbots powered by ChatGPT right on your website.

​EzyCourse​: Create and sell your courses, services, products, or online communities from one platform.

​Katteb: The first fact-checked, real-time, and localized AI writer.

​

What'd you think of today's edition?

• ​🤖🤖🤖🤖🤖 Absolutely Great!​
• ​🤖🤖🤖 Meh, do better​
• ​🤖 You didn't bring the heat​

​