README.md

KAUST Computer Vision Training Program

Slides · GitHub

Program Overview

KAUST Academy's AI Training program offers a comprehensive four-week journey into the world of Computer Vision. In today's rapidly evolving field of computer vision and deep learning, staying at the forefront of research and practical applications is essential. To address the growing demand for expertise in cutting-edge topics, we designed a comprehensive training program that covers a range of critical areas in computer vision and machine learning. This program aims to equip participants from TAHAKOM with the knowledge and hands-on skills required to excel in various domains, including vision-based AI, video understanding, advanced training paradigms, and multi-modal learning.

Course Schedule

Table of Contents

• Week 1: Transformer for Vision

  1. Introduction to Transformers Network
  2. Vision Transformer
  3. Transformers in Computer Vision
  4. Vision Transformers for Downstream Tasks
  5. Advanced Topics and Emerging Trends

• Week 3: Beyond the Supervised Training Paradigm

  1. Introduction to Self-Supervised Learning (SSL)
  2. Advanced SSL Techniques and Applications
  3. Exploring Generalization Abilities of Computer Vision Models
  4. Addressing Data Scarcity in Computer Vision
  5. Improving Computer Vision models efficiency

Getting Started

Prerequisites

conda create --prefix ./venv python=3.10.12
conda activate ./venv
pip install jupyter

Launch the repository on KAUST ClassHub

Use in the public cloud

Week 1: Transformer for Vision

Tutorial	Open in Google Colab	Open in Kaggle
1. Intro: Transformer
2. Vision Transformer
3. Transformer vs CNN
4. DETR

Week 3: Beyond the Supervised Training Paradigm

Tutorial	Open in Google Colab	Open in Kaggle
1.1. Introduction to SSL
1.2. Introduction to SSL
2. MAE
3. Domain Adaptation
4. Data Scarcity
5. Efficient Models