Self-Supervised Machine Learning

Course details

Are you a programmer looking to expand your model building skills in hopes of gathering more information from the data you have available? This course teaches you how self-supervised machine learning can assist you. Instructor Janani Ravi shows you how self-supervised models work and how to leverage self-supervised models in transfer learning to improve the performance of downstream tasks such as image classification.

Concepts

0. Introduction

01 - Course prerequisites
02 - The need for self-supervised learning
03 - Supervised, unsupervised, and self-supervised learning
04 - Constructing self-supervised tasks
05 - Generalized representations and pretext tasks

1. Self-Supervised Learning in Computer Vision

06 - Self-supervised learning in computer vision
07 - Pretext task - Predicting image rotation
08 - Pretext task - Image colorization
09 - Pretext task - Image inpainting
10 - Pretext task - Predictive relative position
11 - Pretext task - Jigsaw task
12 - Pretext task - Video
13 - Pretext task - Video and sound

2. Contrastive and Non-contrastive Learning Techniques

14 - Objective of self-supervised learning
15 - Contrastive and non-contrastive learning
16 - Contrastive learning
17 - Pretext Invariant Representation Learning (PIRL)
18 - Clustering
19 - Distillation
20 - BYOL and SimSiam
21 - Redundancy reduction - Barlow Twins

3. Getting Set Up with the Demo Environment

22 - PyTorch lightning and lightning bolts
23 - Exploring the flowers data set
24 - Using Google Colab

4. Transfer Learning Using Supervised Pretraining

25 - Brief overview of transfer learning
26 - Loading the image data set
27 - Setting up the data module
28 - Setting up the classification model
29 - Training the classification model - Frozen backbone
30 - Training the classification model - Fine-tuning backbone

5. Self-Supervised Contrastive Learning with SimCLR

31 - Contrastive learning - SimCLR
32 - Loading the data set
33 - Image classification using the SimCLR pretrained model

6. Self-Supervised Clustering with SwAV

34 - Non-contrastive learning - Clustering with SwAV
35 - Image classification using the SwAV pretrained model

Conclusion

36 - Summary and next steps