Extensive-Vision-AI-Program (ONGOING)

Extensive Vision AI program is a initiative by The School of AI annd Inkers to provide better understanding about Deep Learning and Computer Vision.

• All the Open-Assignments of Extensive Vision AI Program(EVA) can be found here.

Course Structure (ONGOING)

1. Background & Basics: Machine Learning Intuition, Background & Basics of CNN

2. Neural Architecture: Exhaustive Insights into the Neural Architecture (In classroom Coding or ICC)

3. First Neural Networks: Hands-on (ICC) to custom design a DNN

4. Mathematics: Session on the mathematics behind DNN, loss functions, gradient descent, different initialization types, and optimizations (ICC)

5. Batch Normalization & Regularization: In-depth coverage on Batch Normalization techniques and different kind of Regularizations, including noise robustness (ICC)

6. Advanced Convolutions: Advanced Convolutions & Pooling operations with Code examples and usage(ICC)

7. Receptive Field: Exhaustive Coverage on Receptive Fields, advancements in Receptive Field, and how RF diverges for different kind of problems

8. Data Augmentation Techniques: Advanced Image Augmentation Techniques, benchmarks against different techniques and ICC

9. Kernel/Channel Visualization: The most powerful debugging tool at your disposal! (ICC)

10. Advanced Training Concepts: Advanced concepts on training, including LR, Momentum, Learning Rate Finder,

11. SuperConvergence: Advanced topics cover to understand and execute Super Convergence

12. ResNet Part 1: Understanding ResNet end to end (ICC)

13. ResNext Part 2: Understanding ResNet V2, V3 and ResNext (ICC)

14. Inception Network: Understanding Inception Networks (ICC)

15. DenseNet: Understanding DenseNet and it's applications (ICC)

16. MegaProject: Training ImageNet from scratch with Super Convergence close to StateOfAccuracy

17. Small DNNs & their advantages Part 1: Building SqueezeNet & MobileNet from scratch. Includes Kernel Reduction, Channel Reduction, Evenly Spaced Downsampling, Cardinality, Shuffle Operation

18. Small DNNs & their advantages Part 2: Evenly Spaced Downsampling, Cardinality, Shuffle Operation, Distillation & Compression

19. Transfer Learning: Transfer Learning and approaches. (ICC)

20. YOLO v2: YOLO V2 Architecture and Design Decisions

21. YOLO V2 Training: Training YOLO V2 on a custom dataset (with Transfer Learning)

22. Face Recognition: Building a Face Recognition Model from scratch with advanced Loss functions. ICC

23. FR using Siamese Network: Building an FR model using Siamese Network. ICC

24. Zero & One-shot learning: Building a DNN to detect an unseen or never-trained-on object! ICC

25. UNET: Understanding UNET and it's state of art implementations (image segmentation, etc) ICC

26. eNAS: How to train a neural network to write a state-of-art neural network

27. Encoder Decoder Architecture: Representation Learning, Sequence to Sequence Mapping and ICC

28. GAN & Style Transfer: Generative Adversarial Network and many approaches for the same (DCGAN, CycleGAN). Mode Collapse, Non-convergence and ICC

29. Variational Autoencoders: Latent Representations using Variational Autoencoders. ICC

30. Word2Vec & Neural Word Embeddings: Using Word2Vec, ELMO, BERT, GPT-2, Glove & Doc2Vec. ICC

31. RNN: RNN Basics, advances and drawbacks. Visualizing memorizations in RNNs

32. LSTM & GRU: The intuition behind LSTM and GRUs. ICC

33. Attention Mechanism & Memory Networks: Attention & augmented RNNs. Why "Attention"? Memory Networks and ICC

34. Reinforcement Learning Basics: Background, Intuition, and roadmap

35. RL Common Approaches: Building various deep learning agents including DQN and A3C (ICC)

36. OpenGym & RL Basics: OpenAI GYM, and implementation of Q Learning (ICC)

37. Policy Gradients: Policy Gradient Methods, Continuous Action Spaces, and solving several OpenGym problems (ICC)

38. Deep Q-Learning: Deep Q Learning, Replay Memory, Partially Observable MDPs and ICC

39. A3C in depth: A3C in depth and implementation (ICC)

40. AlphaZero: Training an AlphaZero model from scratch!