You are on page 1of 37

19EEE362

Deep Learning for visual


computing
300 (3) Dr. Lekshmi R. R.
Asst. Prof.
Department of Electrical & Electronics
Engineering
Amrita School of Engineering 1
Briefing
Soft Computing
• An approach to computing.
• Gives solutions to complex real-life problems with
uncertainty .
• Soft computing is tolerant of
– Imprecision
– Uncertainty
– Partial truth
– Approximations
Applications
Weather Forecast
Dish Wash
Time
Clean Dishes

Water level

Dr. Lekshmi R. R. 7
Drip Irrigation System

Time Moisture soil

Dr. Lekshmi R. R. 8
Washing Machine

Time Water level


Washing Machine
Time

Water level Clean Clothes


Robotic arm using PID controller

Dr. Lekshmi R. R. 11
Learning and Recognition
Optimal Routing
Swarm Intelligence
Biological Neuron
Artificial Neuron
𝑋1 𝑤1

𝑋2 𝑤2 𝑚 1 𝑖𝑓 ෍ 𝑥𝑤 ≥ 𝜀
෍ 𝑥𝑖 𝑤𝑖 𝑓 𝑥 = 𝑌
0 𝑖𝑓 ෍ 𝑥𝑤 < 𝜀
𝑖=1
𝑤𝑚
𝑚

𝑋𝑚 𝑦𝑖𝑛 = ෍ 𝑥𝑖 𝑤𝑖 𝑦 = 𝑓 𝑦𝑖𝑛
𝑖=1

Inputs Weights Summation Activation Output


Science which deals with making machine intelligence
Technology that allows computers to perform specific tasks intelligently
Science which deals with making machine intelligence
Seeks to build intelligent systems or machines that can automatically learn and train themselves
through experience, without being explicitly programmed or requiring any human intervention
Research Work
• Soft computing theory and techniques
– first introduced in 1980s
• became major research and study area
– in automatic control engineering.
• With the advent of
– digital processors
• low-cost
• very high performance
– reduction of the cost of memory chips
• techniques and application areas of soft computing continued to
expand.
Research Work
• Significant advances in capabilities of machine learning in recent years
as a result of:
– Technical developments in the field
– Increased availability of data
– Increased computing power
• As a result of these advances:
– systems which only a few years ago struggled to achieve accurate results can now
outperform humans at specific tasks.
• There now exist voice and object recognition systems that can perform
better than humans at certain tasks.
• Researchers
– Exploring ways to make models more flexible
– Seeking techniques that allow a machine to apply context learned from one task to
future, different tasks.
Machine learning Applications

Classification
Machine learning Applications
Regression
Pattern Recognition
Eye detector

* =
Eye detector

* =
Eye detector

* =
Natural Language Processing

Speech recognition Create title


Syllabus
Unit 1
Introduction to Visual Computing and Neural Networks, ANN-Perceptron, Feed
forward network, Back propagation algorithm. Activation functions -linear,
softmax, tanh, ReLU; error functions. Cost functions, Learning Rate Dynamics,
optimization and Computational Complexity. Transfer Learning of a Network.
Unit 2
Introduction to Convolutional Neural Networks (CNN) and LeNet, Convolutional
Autoencoders and Deep CNN (AlexNet, VGGNet), Very Deep CNN for
Classification (GoogLeNet, ResNet, DenseNet). Autoencoders, Autoencoders for
Representation Learning and MLP Initialization, Stacked, Sparse, Denoising
Autoencoders and Ladder Training. Object Localization (RCNN) and Semantic
Segmentation.
Unit 3
Generative Models with Adversarial Learning, Recurrent Neural Networks (RNN)
for Video Classification, LSTM for image/video analysis.
References/Textbooks
1. Goodfellow, Y, Bengio, A. Courville, “Deep
Learning”, MIT Press, 2016.
2. S. Haykin, “Neural Networks and Learning
Machines”,Third Edition,Pearson, 2008
3. Aditi Majumder, M. Gopi, Introduction to
Visual Computing: Core Concepts in
Computer Vision, Graphics, and Image
Processing, CRC Press; 1 edition, 2018.
Course Outcomes

CO Course Outcome

CO01 Understanding of visual computing and Neural networks

CO02 Exposure to convolutional Neural networks and very deep CNN

CO03 Understanding of computational complexity and transfer learning of a


network
CO04 Exposure to Recurrent Neural networks (RNN) for classification

Dr. Lekshmi R. R. 33
CO PO Mapping

PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 PSO1 PSO2

CO01 3

CO02 3

CO03 3 2

C004 3 2 1 1
Course Int:
Type Ext

LTP CA1 CA2 CA3 MT CA4 CA5 ES

60:
Simula Simula
40 Class
tion tion
300 Q1 Exam Project participat Exam
based based
ion
Task_1 Task_2

Marks 5 5 5 50 10 5 100

Weightage 5 5 5 30 10 5 40
Thank you
Notes
• Briefing.doc
• Machine learning.doc

You might also like