کلاس و دوره آموزشی نیمه خصوصی یادگیری عمیق (Deep Learning Fundamentals) با استاد مهدی شکری زاده

1 - Enthtusiastic people who just began their journey in the realm of AI

2 - Professionals who pursue further insight into deep learning

3 - Any student with a desire to become expert

Introduction to Deep Learning
- History
- How 'Deep' is Deep?

Tools of the Trade
- Python, Keras, Mxnet
- What About TensorFlow?
- Do I Need OpenCV?

Image Fundamentals
- Pixels
- Farming Images
- Image Coordinate System
- Images as Numpy Arrays
- RGB or BGR?
- Scaling and Aspect Ratio

Image Classification Basics
- What is Classification?
- Semantics
- Supervised / Unsupervised / Semi-Supervised Learning
- Deep Learning Classification Pipeline
- Gather / Split / Train / Evaluate
- Feature-based Learning Vs. Deep Learning

DataSets for Image Classification
- MNIST
- Animals
- CIFAR-10
- SMILES
- Kaggle
- Flowers-17
- CALTECH-101
- Tiny Imagenet 200
- Adience
- ImageNet
- CVPR
- Stanford Cars
- LISA

Configure Your Own Development Environment

First Image Classifier
- Introducing the Animal Dataset
- Build Your Own Deep Learning Toolkit
- Basic Image PreProcessor
- Building an Image Loader
- KNN
- KNN Hyperparameters
- KNN: Pros and Cons

Parameterized Learning
- The Famous Four
- Linear Classification
- From Images to Labels
- Simple Linear Classifier with Python
- Loss Function
- Multi-Class SVM Loss
- Cross-Entropy Loss
- Softmax Classifiers

Optimizing Methods and Regularization
- Gradient Descent
- Bias trick
- Implementing Basic Gradient Descent in Python
- Stochastic Gradient Descent (SGD)
- Mini-batch SGD
- Extensions of SGD: Momentum, Nesterov
- What is Regularization?
- Updating Loss and Weight with Regularization
- Types of Regularization

Neural Network Fundamentals
- Introduction
- The Famous Perceptron
- Backpropagation and Multi-layet Networks
- Multi-layet Networks with Keras
- The Four Ingredients in a Neural Network
- Weight Initialization
- Constant Initialization
- Uniform and Normal Distributions
- LeCun Uniform and Normal
- Glorat / Xavier Uniform and Normal
- He et. al. / Kaiming / MSRA Uniform and Normal

Convolutional Neural Networks
- Understanding Convolutions
- Convolution Versus Cross-correlation
- The Big and Small Matrices
- Kernels
- Implementing Convolution with Python
- Deep Learning and Convolutions
- CNN Building Blocks
- Layer Types
- Convolutional Layers
- Activation Layers
- Pooling Layers
- Fully-connected Layers
- Batch Normalization
- Dropout
- Common Architectures
- Layer Patterns
- Rules of Thumb
- Translation, Rotation and Scaling in CNNs

Training Your First CNN
- Keras Configuration
- Converting Images to Arrays
- Understanding keras.json File
- Image to Array PreProcessor
- Implementing ShallowNet
- ShallowNet on Animals Dataset
- ShallowNet on CIFAR-10

Saving and Loading Models
- Serializing a Model to Disk
- Loading a Pre-trained Model

LeNet: Recognizing Handwritten Digits
- The LeNet Architecture
- Implementing LeNet
- LeNet on MNIST

MiniVGGNet: Going Deeper with CNNs
- The VGG Family
- The Mini VGGNet Architecture
- Implementing MiniVGGNet
- MiniVGGNet on CIFAR-10
- With and Without Batch Normalization

Learning Rate Schedulers
- Dropping the Learning Rate
- The Standard Decay Schedule in Keras
- Step-based Decay
- Implementing Custom Learning Rate in Keras

Spotting Underfitting and Overfitting
- What are Underfitting and Overfitting?
- Effects of Learning Rates
- Pay Attention to Training Curve
- Validation Loss is Lower Than Training Loss?
- Monitoring the Training Process
- Creating a Training Monitor
- Babysit Your Trainer

Checkpointing Models
- Checkpointing Neural Network Models
- Improvements
- Checkpointing Best Models

Visualizing Network Architecture
- Impoprtance of Visualization
- Installing graphviz and paydot
- Visualizing Keras Networks

Out-of-the-box CNNs for Classification
- State-of-the-art CNNs in Keras
- VGG16 and VGG19
- ResNet
- Inception V3
- Xception
- Pre-trained ImageNet CNNs

Case Study: Breaking Captchas
Case Study: Smile Detection