Autoplay
Autocomplete
Previous Lesson
Complete and Continue
SigmaML: Deep Learning, Computer Vision and NLP
Welcome
Platform Walkthrough (20:53)
22 Jul - Orientation (37:08)
Schedule
Doubts Form
Gamification
Leaderboard
Daily Streak
Milestone 0 - Recap & Pre-requisites
1 - Calculus (25:07)
2 - Basics of Gradient Descent (62:56)
3 - Coding Gradient Descent and Understanding Learning Rates (33:46)
4 - Versions of GD, Minimas, Initialisation, Feature Scaling, Learning Rate (43:56)
5 - Why Tensorflow? (31:34)
6 - Project: Intro to Object Segmentation (52:52)
7 - Project: Preprocessing (30:56)
8 - Project: Model Building and Evaluation (63:23)
Refresher Quiz (Non-graded)
Milestone 1 - Intro to Tensorflow & Professional Workflows
Day 1 - Tensorboards: What and Why? (22:26)
Day 2 - Tensorboards: Default Callback Logging + Custom Image, Scalar Logging, and more! (52:21)
Day 3 (a) - Saving Models (31:08)
Day 3 (b) - Moving to a professional file structure (37:02)
Day 4 - Quiz 1
Milestone 2 - Your First Neural Network!
Day 1 (a) - Intro to DL, Perceptron, MLP (34:50)
Day 1 (b) - Deep Learning v/s Classical ML (28:44)
Day 2 - How NN learn, Notation, Forward Pass (45:51)
Day 3 - Trainable Parameters, Backpropagation Derivation, and Memoization (40:36)
Day 4 - Let's code it up! (40:49)
Day 5 - What are the correct number of epochs? (30:10)
Day 6 & 7 - Quiz 2
Day 8 - Intro to Activation Functions & The Important of Non-linearity (41:02)
Day 9 (a) - Different Activation Functions with properties (33:58)
Day 9 (b) - Dying ReLU: what, why, and the solution! (23:33)
Day 9 (c) - Variants of ReLU, SoftMax (16:17)
Day 10 - Loss Functions (56:58)
Day 11 - Recap of Information Theory (79:00)
Milestone 3 - Practical Considerations (Phase A)
Day 1 - What is the correct batch size? (43:22)
Day 2 - Vanishing Gradient: Problem (35:31)
Day 3 - (a) Vanishing Gradient: Detection & Solution (14:47)
Day 3 - (b) Exploding Gradient Problem & Gradient Clipping (20:18)
Day 4 - (a) Weight Initialisation (22:31)
Day 4 - (b) Code: Activation Fns, Vanishing/Exploding Grad, Weight Init (29:29)
Day 5 - Depth v/s Width (18:46)
Milestone 4 - Dealing with Images
Day 1 - Color Spaces and Types of Images (47:41)
Day 2 - Image Manipulation with Filters (53:55)
Day 3 - OpenCV: Color Spaces, Equivalisation, and Thresholding (45:12)
Day 4 - OpenCV: Image Smoothing, Morphological Ops, Edge Detection (26:46)
Day 5 - Quiz 3
Day 6 - How CNNs were built? (31:20)
Day 7 (a) - Convolution, Feature Maps, and Output Sizes (35:54)
Day 7 (b) - Parameter Sharing, Num of Params, Receptive Field (23:58)
Day 8 (a) - Visualising the feature maps (26:30)
Day 8 (b) - Let's play pool! (32:45)
Day 9 - Coding a CNN, CNN v/s ANN on Cifar-10, Softmax Stability, Evolution of CNNs (LeNet, AlexNet, etc.) (61:54)
Day 10 & 11 - Model Building: Pre-existing, Pre-training, Transfer, Finetune, Auxilliary, Unsupervised Pre-training (75:39)
Day 12 & 13 - Quiz 4: Build a CNN on the Cifar-100 dataset
Day 14 - Data Augmentation (40:24)
Day 15 - Overfitting and Dropout (62:14)
Day 16 (a) - Implementing Dropout (8:29)
Day 16 (b) - Regularisation in Neural Networks (12:07)
Day 17 - ResNet, Residual Blocks, and Skip Connections (34:26)
Day 18 - Encoder-Decoder, Convolution Transpose, U-net (42:49)
Milestone 5 - Hackathon
Intro to Kaggle (26:35)
19 Sep - Intro Call (43:41)
Template Notebook
Quiz 5: Submit Your Github Repo for the Hackathon
Milestone 6 - Special Neural Networks
Day 1 - Introduction to Autoencoders & their types (47:54)
Day 2 (a) - Implementing Auto-Encoders, with Application (Noise Removal - DAE, Anomaly Detection) (38:34)
Day 2 (b) - Unsupervised Pre-training (9:55)
Day 3 - Implementing Self-supervised Learning Models: Super Resolution + Inpainting (41:34)
Day 4 - Siamese Network & One-shot learning (56:23)
Day 5 - Siamese in 3D + Retrieval Problems + Embedding Space (32:33)
Day 6 - Siamese Network Implementation (27:41)
Day 7 - Multi-task learning & Multiple Heads (34:59)
Day 8 - Generative Modeling - Intro to VAEs (16:53)
Milestone 7 - Dealing with Text
Day 1 - Text Processing: Built-in Tokenisers and Training Custom Tokenisers using HuggingFace (43:52)
Day 2 (a): Text Normalisation: Stemming, Stopwords, Lemmatisation (17:54)
Day 2 (b): POS Tagger, NER (using NLTK & SpaCy) (16:43)
Day 3 (a): Text to Vectors: BOW, N-grams, TF-IDF (45:47)
Day 3 (b) - Text to Vectors: Implementation (using Sklearn) (12:19)
Day 4 - Learned Word Embeddings: Word2Vec (CBOW & Skip-gram) (59:30)
Day 5 (a): Pre-trained Word2Vec and Custom Training FastText (using Gensim) (14:42)
Day 5 (b): RNNs - Premise, Arch, Unfolding, Forward Prop, Visualisation (37:22)
Day 6 (a): RNNs - Back-prop, Issues, Variants (Stacked, Bi-directional) (36:33)
Day 6 (b): Implementing RNNs and LSTMs (44:59)
Day 7: LSTM, GRU
Day 8 & 9: Quiz 5 - Identify the Language behind the Question!
Day 10: Seq-to-Seq, Attention, and the LLM story (50:32)
Day 11: Attention: Math, Models, and Types (39:04)
Milestone 8 - Projects
1.1 - Building a Conversational AI (21 Oct) (59:34)
Template Slides for Demo Day
Quiz 6: Demo Day Submissions
GenAI Specials
Day 9 - GANs
Practical Considerations (Phase B)
Optimizers
Niche Modalities
3D Data
Support Sessions
28 Apr - Orientation (31:24)
1.1 - 12 May - Tensorflow (27:04)
5.1 - 29 Sep - Hackathon Review (112:02)
Extras
1.1 - Intro to Prompt Engineering (93:41)
2.1 - 4 Aug - Modern Way of Building AI systems (69:21)
3.1 - Priciples of MLOps (121:16)
Reading Room & Labs
Seminal Papers List
1.1: Introduction to Research in AI (111:53)
1.2: Submit Your Literature Survey!
2.1: 9 June - Live Session on Branding Beyond Github Profiles! (28:11)
2.2: Submit Your Medium Blog!
3.1 - Lab Session: Let's Build A Virtual TryOn! (Model Designs, Inpainting, Running GitHub Repos, Custom Data Loaders & Losses) (207:54)
Day 10: Seq-to-Seq, Attention, and the LLM story
Lesson content locked
If you're already enrolled,
you'll need to login
.
Enroll in Course to Unlock