Extensive Vision AI Program

Phase 1:

• Background & Basics: Machine Learning Intuition
• Python: Python 101 for Machine Learning
• DNN Concepts: Convolutions, Pooling Operations & Channels
• PyTorch: Pytorch 101 for Vision Machine Learning
• First Neural Network: Kernels, Activations, and Layers
• Architectural Basics: We go through 9 model iterations together, step-by-step to find the final architecture
• Receptive Fields: The CORE fundamental concept behind EVA Program
• BN, Kernels & Regularization: Mathematics behind Batch Normalization, Kernel Initialization, and Regularization
• Backpropagation and Advance Convolutions: Depthwise, Pixel Shuffle, Dilated, Transpose and others
• Advanced Image Augmentation Techniques: Albumentations, Richman's data augmentation and benchmarks
• DNN Interpretability: Class Activation Maps, the most powerful debugging tool at your disposal
• Advanced Training Concepts: Optimizers, LR Schedules, LR Finder & Loss Functions
• SuperConvergence: Cyclic Learning Rates, One Cycle Policy, and Dawnbench
• ResNets: Training ResNet for TinyImageNet from scratch
• Inception: Understanding Inception and DenseNet Architectures
• YoloV2: Understanding YOLOV2 Loss Function
• YoloV5: Implementing Object Detection Training & Transfer Learning
• RCNN Family: RCNN, Fast-RCNN, FasterRCNN & MaskRCNN
• CapStone: Monocular Depth Estimation and Background/Foreground Extraction

Phase 2

• Deploying over AWS: Train, Dockerize and then deploy your model on AWS.
• MobileNet & Other Edge DNNs: Training a DNN for EDGE Deployment from scratch. Understanding MobileNets and ShuffleNets
• Face Recognition Part 1: Face Detection and Detection Strategies
• OpenCV Refresher and Face Recognition Part 2: Implementing Object Tracking and Stabilization, OpenCV and DLIB, for face recognition and others
• Human Pose Estimation: State of Art HPE and Human Localization
• Super-Resolution/Neural-Style-Transfer: Leveraging Transfer learning for NST and "Dense" models
• Segmentation and Usage in Medical Domain: U-NET, its relatives and usage in Medical Science
• GANs: Generative Adversarial Network and Variants
• Encoders: Auto Encoders and Variational AutoEncoders
• Neural Work Embedding: The Embedding Layer
• Sequence Models: RNNs and LSTMs
• GRU, Attention Mechanism & Transformers: Attention is all you need!

Phase 3

• Reinforcement Learning Basics: Markov Decision Processes, Deterministic, and Stochastic Environments & Bellman Equation
• Q-Learning: Q-Learning, Plan vs Policy Networks, and Environment Models
• Deep Q-Learning & DeepTraffic: Custom Environments, OpenGym, Exploration vs Exploitation, and improvements to DQN
• Deep Reinforcement Learning: Policy Gradients, Dynamic Programming, Policy Evaluations, and Temporal Difference Learning
• Actor-Critic Models: Memory Structures, Gibbs Softmax, Eligibility Traces, and Polyak Averaging
• A3C Models: A3C, A3C optimizations,, and implementation logic
• Deep Deterministic Policy Gradients: DDPG Background, Off-Policy Networks, Continuous Action Spaces, and Replay Buffers
• Twin Delayed DDPG Part 1: Clipped Double-Q Learning, Delayed Policy Updates, and Target Policy Smoothing
• Twin Delayed DDPG Part 2: Full TD3 implementation to make a robot walk, and solve a custom environment