Image Segmentation Prompt

'

Weakly Supervised Image Prompt Segmentation with Foundation Models documentation!

'

Introduction

• Project Introduction
  • image segmentation prompt
  • Documentation axes

'

Building Blocks

• PyTorch
  • 1. Introduction
  • 2. Installing PyTorch with Anaconda
  • 3. Introduction to Tensors
  • 4. Datasets & DataLoaders
• Neural Network
  • 1. Introduction
  • 2. dataset:
  • 3. Training on a simple dataset
  • 4. Test the DataPreprocessing class
  • 5. test the DataExploration class:
  • 5. test the NeuralNetwork class
  • 6. Testing the ModelTraining class
  • 7. test the ModelEvaluation class
  • Link to github repository and colab applications:
• Transformer Architecture
  • 1. General Introduction
    • a. Overview
    • b. The purpose of Transformer networks
    • c. The Transformer Architecture
    • d. Key Components
  • 2. The Encoder
    • a. Tokenizer
    • b. Input embedding
    • c. Positional Encoding
    • d. self Attention
    • e. Multi-Head Attention
    • f. Add and norm - Norm
    • g. Feed Forward
    • h. Residual Connections
    • i. Conclusion
    • j. BIBLIOGRAPHIC
  • 3. The Decoder
    • a. Introduction
    • b. difference between decoder and encoder
    • c. Masked multi-head attention
    • d. Multi-Head Attention
    • e. Feed Forward
    • f. Conclusion
  • 3. Softmax in Transformers
• Visual Transformer (ViT)
  • 1. introduction
  • 2. Vision Transformers example
• ViT project implemention
  • Table of Contents
  • 1. Vision Transformer - Pytorch
  • 2. Install
  • 3. Dataset preparation
    • Classification dataset
    • Datasets_DataLoaders
    • Resize_Tensorize
    • Sample_Viz
  • 4. ViT components
    • PatchEmbedding
    • Positional Embedding
    • MultiheadSelfAttentionBlock
    • MLPBlock
    • TransformerEncoderBlock
    • ViT
  • 5. code source

'

Foundation Models

• Definition
  • Introduction to Foundation models
  • What are Foundation Models ?
  • History of Foundation Models
  • Types of Foundation Models
  • Applications of Foundation Models
  • Limitations of Foundation Models
  • Conclusion
• Famous Models
  • Grounding DINO
    • 1. Introduction
    • 2. Grounding DINO Performance
    • 3. Advantages of Grounding DINO
    • 4. Grounding DINO Architecture
    • 5. Reference
  • Segment Anyting Model
    • 1. Introduction to SAM:
    • 2. What is the Segment Anything Model?
    • 3. SAM’s network architecture
    • 4. How does SAM support real-life cases?
    • 5. Reference
• Groundingsam project implemention
  • Cloning the Repository
  • Setting the Working Directory
  • Navigating to the Directory
  • Installing Dependencies
  • Creating Directories for Models and Annotations
  • Downloading GroundingDINO Model Weights
  • Returning to the Main Directory
  • Installing Segment Anything via pip
  • Downloading SAM Model Weights
  • Returning to the Main Directory (Again)
  • Importing and Initializing
  • Detection, Annotation, and Segmentation
  • Summary

'

image Segmentation

• State Of The Art
• Prompt Generator / Analyzer
  • Overview
  • Prompt Gemini Generator
  • Features
    • Prompt pre-processing
    • Similarity Reduction
    • Complexity Analysis
    • Readability Analysis
  • Prompt processing
  • Prompt Generator Example
  • Prompt Analyzer Exemple

'

References

• Transformer: Attention Is All You Need
  • 1. Introduction
  • 2. Background
  • 3. Self-Attention Mechanism
  • 4. Multi-Head Self-Attention
  • 5. Position-Wise Feed-Forward Networks
  • 6. Transformer Model
  • 7. Attention Visualization
  • 8. Experimental Results
  • 9. Conclusion
  • Summary
• AN IMAGE IS WORTH 16X16 WORDS
  • 1. Objectives of the Paper
  • 2. Paper Contributions
  • 3. Paper Limitations, Further Research
  • Summary
• PASCAL VOC
  • 1. Introduction
  • 2. One annotation:
  • 3. Multiple annotations:
  • 4. Additional fields:
  • 5. Conclusion
• Grounding DINO
  • Introduction:
  • Main Concept:
  • Advantages:
  • Existing Approaches:
  • Performance:
  • Contributions:
  • General Conclusion of the Paper
• Segment Anything
  • Introduction
  • Key Components
    • 1. Task: Promptable Segmentation
    • 2. Model: Segment Anything Model (SAM)
    • 3. Data: SA-1B Dataset
  • Methodology
  • Experiments and Results
  • Conclusion

'