This course is a suitable choice for beginners in Deep Learning and Computer Vision. It encompasses the rudiments of Neural Networks, Convolutions and the essential math concepts with a real-world case study.

Objective

The aim is to share knowledge in the field of Computer Vision with the assistance of an end-to-end pipeline starting with Data processing till preparing a deep learning model for edge deployment. 

• How artificial Neuron functions and a Neural Network is formed
• Understanding the learning  process of Neural Networks
• Optimization of  model training
• The idea of Pre-trained models and Transfer Learning
• Implementation of an Image Classifier in keras(Tensorflow) and preparing the model for edge deployment

• A complete understanding of the entire learning process in CNN
• How to preprocess the images to increase the training samples
• Good knowledge of model implementation in keras(tensorflow)
• Entire pipeline understanding for edge device deployment

Python basics(nice to have)

Session 1: Introduction to Deep Learning

• Basics of Machine Learning & Deep Learning
• Why Deep Learning
• Linear and Nonlinear functions
• What are learnable parameters?
• Forward propagation and Backward pass in DL
• Loss function and its significance
• Implementation of a simple Neural Network

Session 2: Enhancing the Learning process

• Bias-Variance Tradeoff
• What is Underfitting and Overfitting
• Various optimizers for deep learning
• Implementation of an Image Classifier using FNN(keras)

Session 3: Convolution Neural Networks

• What are CNNs and Why they are introduced
• Filters, Channels, Pooling layers
• Learning process in CNNs
• Data Augmentation
• Pretrained Models
• Transfer Learning
• Implementation of an Image Classifier using CNN(keras)

Session 4: Real-world case study (Edge Deployment)

• What are edge devices
• Model Quantization and the significance
• Different Quantization Techniques
• Building an Image Classifier to Identify disease in the plants
• Model Conversion to TFlite format
• Edge Compiler and quantized model