Modern Deep Convolutional Neural Networks with PyTorch

Image Recognition with Convolutional Neural Networks. Advanced techniques for Deep Learning and Representation learning

Created byDenis Volkhonskiy

Last updated 2/2020

English

What you'll learn

Convolutional Neural Networks
Image Processing
Advance Deep Learning Techniques
Regularization, Normalization
Transfer Learning

Course content

5 sections • 29 lectures • 1h 54m total length

Introduction3:39
Computer Vision Problems2:32
Linear Layer and Classification Pipeline4:39
Loss functions and Softmax3:49
Stochastic Gradient Descend5:08
PRACTICE #1: Data loading4:48
Load and preprocess the CFR 10-class dataset in pytorch, applying tensor conversion and normalization with mean 0.5 and std 0.5, then batch 4 with 2 workers for a linear classifier.
PRACTICE #2: Linear Classifier in PyTorch (part 1)6:15
PRACTICE #3: Linear Classifier in PyTorch (part 2)3:18
PRACTICE #4: Multi-layer perceptron5:34

What is image1:57
Defines what an image is, explains grayscale and color channel representations, and contrasts hand-crafted image descriptors with deep learning's automatic feature learning in neural networks.
Motivation to Convolutions1:32
Convolution operation2:30
Parameters of the convolution2:42
Non-linear function2:06
Max Pooling and Average Pooling1:50
Building deep convolutional network1:10
PRACTICE #5: Convolutional Neural Network10:17

Overfitting. L2 regularization3:08
DropOut regularization. DropConnect regularization4:20
Explore dropout and dropconnect regularization as deep learning techniques that randomly zero features or weights during training, and understand training versus testing, Bernoulli masking, and tuning the dropout probability.
DropBlock regularization4:17
Explore dropblock regularization for deep convolutional networks, learning to zero random image blocks during training, improving classification accuracy and including block size, masking, and class activation mapping visualizations.
Early Stopping regularization3:23
Batch Normalization5:48
Understand batch normalization, which speeds training by normalizing data to zero mean and unit variance using feature-wise statistics, epsilon, and learnable parameters, with comparisons to layer normalization.

Requirements

Machine Learning
Linear Regression and Classification
Matrix Calculus, Probability
Deep Learning basis: Multi perceptron, optimization
Python, PyTorch

Description

Dear friend, welcome to the course "Modern Deep Convolutional Neural Networks"! I tried to do my best in order to share my practical experience in Deep Learning and Computer vision with you.

The course consists of 4 blocks:

Introduction section, where I remind you, what is Linear layers, SGD, and how to train Deep Networks.
Convolution section, where we discuss convolutions, it's parameters, advantages and disadvantages.
Regularization and normalization section, where I share with you useful tips and tricks in Deep Learning.
Fine tuning, transfer learning, modern datasets and architectures

If you don't understand something, feel free to ask equations. I will answer you directly or will make a video explanation.

Prerequisites:

Matrix calculus, Linear Algebra, Probability theory and Statistics
Basics of Machine Learning: Regularization, Linear Regression and Classification,
Basics of Deep Learning: Linear layers, SGD, Multi-layer perceptron
Python, Basics of PyTorch

Who this course is for:

Who knows a bit about neural networks
Who wants to enrich their Deep Learning and Image Processing knowledge
Who wants to study advanced techniques and practices

Modern Deep Convolutional Neural Networks with PyTorch

What you'll learn

Explore related topics

Course content

Introduction9 lectures • 40min

Convolutional Neural Networks8 lectures • 24min

Regularization and Normalization5 lectures • 21min

Improving the quality4 lectures • 17min

Boat Recognition Project3 lectures • 14min

Requirements

Description

Who this course is for: