Advanced Computer Vision Replearning, Vae, Gan, Deepfake +

Advanced CV Deep Representation Learning, Transformer, Data Augmentation VAE, GAN, DEEPFAKE +More in Pytorch & Numpy

What you'll learn

Representation Learning

Deep Unsupervised/Supervised/Self Supervised Visual Representation Learning Techniques

Industry Level Advanced Computer Vision

Awesome SOTA Data Augmentation techniques in pytorch

Various properties of Softmax and CrossEntropy in Numpy & Pytorch

State of the art methods like RandAug, JigSaw, PEARL, NPILD, SimCLR, SupCon and many more..

SimCLR (Simple Contrastive Learning), Supervised contrastive learning

Faiss Search, Image Search and Cluster Search

noise contrastive estimator

Visual Transformers

AutoEncoders, VAE, GAN

DeepFake

Requirements

Desire to learn something awesome and new!

Description

Published in  2021: Alpha ReleaseYou can take this course risk-free and if you don't like it, you can get a refund anytime in the first 30 days!Welcome to the "Advanced CV Deep Representation Learning, Transformer, Data Augmentation VAE, GAN, DEEPFAKE +More in Pytorch & Numpy".Deep Unsupervised Visual Representation Learning, Unsupervised computer vision in deep learning is very niche skill and it is being heavily used in production by AI superstar companies like Google, Amazon, Facebook, as a matter of fact lots of ideas we will talk about. In this course are being used to build SOTA products like Shop the Look or Face Search, Speech to emotion detection.To learn Deep Learning and Deep Unsupervised Visual Representation learning,  step-by-step, you have come to the right place! ===============================Deep Learning is Easy to learn, if you know basic Math and can code..Thanks to my several years of experience in Deep Learning, I wanted to share my experience in Deep Representation Learning which are highly used in production level applications.We'll take a step-by-step approach to learn all the fundamentals of Representation learning, Various kind of Visual Representation learning, SOTA data augmentations,  .At the end of this course, you'll be productive and you'll know the following:First PartUnsupervised Visual Representation learningNumpypytorchpytorch Tensor APIpytorch Tensor Manipulationpytorch Autograds and gradientspytorch Vision training pipelinetorchvision pretrained model loadImage SearchCluster SearchFaiss SearchPEARL NPILDJigSaw Simple Contrastive learningSupervised Contrastive learningSelf Supervised Contrastive learningPart 2VAEGANDEEPFakeNote:  The Hands on section is written in python 3.6, pytorch, numpy which is defacto now a days for deep learning. But the concepts covered in the course is also applicable if you use tensorflow or other equivalent libraries.Although the code is Computer Vision heavy but these ideas can also be applied to Speech and NLP.===============================You can take this course risk-free and if you don't like it, you can get a refund anytime in the first 30 days!===============================InstructorThe instructor of this course have more than 15+ years of experience in Machine learning and deep Learning, and worked with people from Google Brain team. The instructor also hold multiple patent in the area of machine learning and deep learning. Fish AI is in stealth mode early stage start up as of 2021.===============================This Course Also Comes With:Lifetime Access to All Future UpdatesA responsive instructor in the Q&A SectionLinks to interesting articles, and lots of good code to base your next applications ontoUdemy Certificate of Completion Ready for DownloadThis is the course that could improve your career!Computer vision is a niche skill. Especially if you know deep learning unsupervised approches. All the papers and ideas presented in this course are used by production level AI products. the skills you acquire in this course will definitely help you in lots of computer vision applications.I hope to see you inside the course!Who this course is for:AI application Developers who want to built cool vision based applicationsAI application Developers who want to learn unsupervised way of deep learning Any Developers who wants to build face recognition, object detection, image search , apparel recognition, speech recognition based productsAI Architects who want to develop state of the art vision productsAnyone looking to learn the theory of deep unsupervised visual representation learningHappy learning!

Overview

Section 1: Introduction

Lecture 1 Course Overview

Lecture 2 Applications

Lecture 3 Google Colab Setup

Lecture 4 Course Structure & Important Notes

Section 2: Data Science in Numpy & Pytorch (code) - Background

Lecture 5 Data Science in Numpy - Part1 (Code)

Lecture 6 Data Science in Pytorch - Part1 (Code)

Lecture 7 Data Science in Pytorch - Part 2(Code)

Section 3: Pytorch AutoGrad

Lecture 8 Pytorch AutoGrad

Lecture 9 Custom CNN in Pytorch

Section 4: Faiss & Image Search (Hands on, Dont skip)

Lecture 10 Image Search(Basic & Cluster)

Lecture 11 Faiss Overview

Lecture 12 Basic Image Search (Code)

Lecture 13 Basic Image Search With pertained Resnet (cifar-10 dataset) (Code)

Lecture 14 Cluster Search (Code)

Section 5: SOTA Data augmentation (Hands On)

Lecture 15 Why Data Augmentation & History

Lecture 16 CutMix Paper Overview

Lecture 17 Results of CutMix

Lecture 18 CutMix Algorithm

Lecture 19 CutMix (Code)

Lecture 20 RandAugment

Lecture 21 RandAugment (Code)

Section 6: Softmax think out of the box (Hands On)

Lecture 22 SoftMax Think out of the box

Lecture 23 Temperature Scaling & soft softmax (code)

Lecture 24 Summery

Section 7: Prelearing & UVR by Context Prediction (Theory)

Lecture 25 Pretext Task

Lecture 26 Overview of Unsupervised Visual Representation Learning by Context Prediction

Lecture 27 Results of UVR by Context Prediction

Section 8: JigSaw

Lecture 28 Overview of Jigsaw

Lecture 29 Network and Training process

Lecture 30 Results of JigSaw

Section 9: Non-Parametric Instance Level Discrimination(NPILD) (hands on)

Lecture 31 Non-Parametric Instance-level Discrimination & Metric learning approach

Lecture 32 NPILD Training Process

Lecture 33 Non Parametric Softmax

Lecture 34 Noise contrastive estimation (NCE) - Part 1

Lecture 35 FULL NCE Loss

Lecture 36 NPILD Put it all together

Lecture 37 NPILD Result

Lecture 38 Non Parametric Softmax (CrossEntropy) (Code)

Section 10: PEARL

Lecture 39 Self-Supervised Learning of Pretext-Invariant Representations (PEARL) - Part 1

Lecture 40 PEARL Overview Part 2

Lecture 41 PEARL Loss

Lecture 42 PEARL Results

Section 11: PEARL and NPILD (code)

Lecture 43 NCE & Memory Bank (Code)

Lecture 44 Network and Training NPILD & Pearl (Code)

Section 12: SimCLR

Lecture 45 SIMCLR Overview

Lecture 46 SIMCLR & Multiview Batch

Lecture 47 SimCLR Algorithm and Loss

Lecture 48 Training Details

Lecture 49 Softmax is invariant under translation (Important)

Section 13: SupCon & SimCLR (Code)

Lecture 50 Supervised Contrastive Learning

Lecture 51 Mocking SimCLR(Code)

Lecture 52 SimClr and Supervised Contrastive Learning (Code)

Section 14: Practice Test (Covering Upto DUVRL)

Section 15: Few More ideas in Visual Representation Learning

Lecture 53 Vissl & Albumentations

Lecture 54 Tips From My Expeience

Lecture 55 Few More ideas

Section 16: DeepFakes & Beyond - Second Part of the course(In-Progress)

Lecture 56 Introduction to DeepFake & Beyond

Lecture 57 Generative Vs Discriminative AI With VAE Example (will be separate course)

Developer who are interested in building AI/Deep Learning products,Architects who are interested in building AI//Deep Learning products,Developer and AI Developer who are interested in Data Augmentation Technique,Developer and AI Developer who are interested in Computer Vision, Deep Learning, Deep Unsupervised Learning