Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Review for
Problem Set 1
CS 231a Winter 201516
Sumant Sharma
Webmaster & SCPD CA
Ph.D. Student, Space Rendezvous Laboratory
Some slides in the presentation are courtesy of
Saumitro Dasgupta (CS 231a CA Winter 201415)
Todays Agenda
Camera Model
Rotation
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Gradient checks
Sanity checks
Babysitting the learning process
Loss function
Train/val accuracy
Weights:Updates ratio
Activation/Gradient distributions per layer
Visualization
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Quick intro without brain analogies
Modeling one neuron
Biological motivation and connections
Single neuron as a linear classi er
Commonly used activation functions
Neural Netw
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Generating some data
Training a Softmax Linear Classi er
Initialize the parameters
Compute the class scores
Compute the loss
Computing the analytic gradient with backpropagatio
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Introduction
Visualizing the loss function
Optimization
Strategy #1: Random Search
Strategy #2: Random Local Search
Strategy #3: Following the gradient
Computing the gradient
N
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Derivatives, Backpropagation, and Vectorization
Justin Johnson
June 1, 2017
1
Derivatives
1.1
Scalar Case
You are probably familiar with the concept of a derivative in the scalar case:
given a function f : R R, the derivative of f at a point x R is define
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
(this page is currently in draft form)
Visualizing what ConvNets learn
Several approaches for understanding and visualizing Convolutional Networks have been developed in the
literature, partly as
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Backpropagation for a Linear Layer
Justin Johnson
April 19, 2017
In these notes we will explicitly derive the equations to use when backpropagating through a linear layer, using minibatches.
During the forward pass, the linear layer takes an input X of sh
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Setting up the data and the model
Data Preprocessing
Weight Initialization
Batch Normalization
Regularization (L2/L1/Maxnorm/Dropout)
Loss functions
Summary
Setting up the data
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Noname manuscript No.
(will be inserted by the editor)
ImageNet Large Scale Visual Recognition Challenge
arXiv:1409.0575v3 [cs.CV] 30 Jan 2015
Olga Russakovsky* Jia Deng* Hao Su Jonathan Krause
Sanjeev Satheesh Sean Ma Zhiheng Huang Andrej Karpathy
Adit
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Understanding Deep Image Representations by Inverting Them
Aravindh Mahendran
University of Oxford
Andrea Vedaldi
University of Oxford
arXiv:1412.0035v1 [cs.CV] 26 Nov 2014
Abstract
Image representations, from SIFT and Bag of Visual
Words to Convolutional
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Do Convnets Learn Correspondence?
Jonathan Long
Ning Zhang
Trevor Darrell
University of California Berkeley
cfw_jonlong, nzhang, [email protected]
Abstract
Convolutional neural nets (convnets) trained from massive labeled datasets [1]
have substantia
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Published as a conference paper at ICLR 2015
E XPLAINING AND H ARNESSING
A DVERSARIAL E XAMPLES
arXiv:1412.6572v3 [stat.ML] 20 Mar 2015
Ian J. Goodfellow, Jonathon Shlens & Christian Szegedy
Google Inc., Mountain View, CA
cfw_goodfellow,shlens,[email protected]
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Andrej Karpathy blog
About
Hacker's guide to Neural Networks
What I learned from competing against a ConvNet on
ImageNet
Sep 2, 2014
The results of the 2014 ImageNet Large Scale Visual Recognition Challenge (ILSVRC) were published a few
days ago. The New
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
arXiv:1312.6034v2 [cs.CV] 19 Apr 2014
Deep Inside Convolutional Networks: Visualising
Image Classification Models and Saliency Maps
Karen Simonyan
Andrea Vedaldi
Andrew Zisserman
Visual Geometry Group, University of Oxford
cfw_karen,vedaldi,[email protected]
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Python Numpy Tutorial
This tutorial was contributed by Justin Johnson.
We will use the Python programming language for all assignments in this course. Python is a great generalpurpose programming
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
Table of Contents:
Intro to Linear classi cation
Linear score function
Interpreting a linear classi er
Loss function
Multiclass SVM
Softmax classi er
SVM vs Softmax
Interactive Web Demo of Linear
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231n Convolutional Neural Networks for Visual Recognition
(These notes are currently in draft form and under development)
Table of Contents:
Transfer Learning
Additional References
Transfer Learning
In practice, very few people train an entire Convoluti
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 14:
Reinforcement Learning
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 14  1
May 23, 2017
Administrative
Grades:
 Midterm grades released last night, see Piazza for more
information and statistics
 A2 and milestone grades scheduled for l
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Speech and Language Processing. Daniel Jurafsky & James H. Martin.
rights reserved. Draft of November 7, 2016.
c 2016.
Copyright
All
CHAPTER
16
Semantics with Dense Vectors
In the previous chapter we saw how to represent a word as a sparse vector with
di
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS231A Midterm
Review
Friday 5/6/2016
Outline
General Logistics
Camera Models
Nonperspective cameras
Calibration
Single View Metrology
Epipolar Geometry
Structure from Motion
Active Stereo and Volumetric Stereo
Fitting and Matching
RANSAC
Hough Transfo
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS 231A Computer Vision, Spring 2016
Problem Set 1
Due Date: April 15th 2016 11:59 pm
1
Projective Geometry Problems [10 points]
In this question, we will examine properties of projective transformations. We define a camera
coordinate system, which is onl
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS 231A Computer Vision, Spring 2016
Problem Set 4
Due Date: May 20th 2016 11:59 pm
1
Viewpoint Estimation With Bag of Words
In the previous problem set, we explored object detection with histogram of oriented gradients
and support vector machines. Often,
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS 231A Computer Vision (Spring 2016)
Problem Set 3
Due: May 4th, 2016 (11:59pm)
1
Space Carving (25 points)
Dense 3D reconstruction is a difficult problem, as tackling it from the Structure from Motion
framework (as seen in the previous problem set) requ
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
PS2 Review
CS231A
Computer Vision: From 3D Reconstruction to Recognition
Spring 2016
1
Outline
Q2: Estimating Fundamental Matrix
Q3: Affine Structure From Motion
Q4: Projective Triangulation in Structure From Motion
2
Least Squares Eight Point Algorithm
I
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
CS 231A Computer Vision (Winter 2016)
Problem Set 2
April 25th 2016 11:59pm
1
Fundamental Matrix (20 points)
In this question, you will explore some properties of fundamental matrix and derive a minimal
parameterization for it.
a Show that two 3 4 camera
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 6:
Training Neural Networks,
Part I
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 6  1
April 20, 2017
Administrative
Assignment 1 due Thursday (today), 11:59pm on Canvas
Assignment 2 out today
Project proposal due Tuesday April 25
Notes on b
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 9:
CNN Architectures
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 9  1
May 2, 2017
Administrative
A2 due Thu May 4
Midterm: Inclass Tue May 9. Covers material through Thu
May 4 lecture.
Poster session: Tue June 6, 123pm
FeiFei Li & Justi
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Adversarial Examples
and Adversarial Training
Ian Goodfellow, Sta Research Scientist, Google Brain
CS 231n, Stanford University, 20170530
Overview
What are adversarial examples?
Why do they happen?
How can they be used to compromise machine learning
sys
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 10:
Recurrent Neural Networks
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 10 
1 May 4, 2017
Administrative
A1 grades will go out soon
A2 is due today (11:59pm)
Midterm is inclass on Tuesday!
We will send out details on where to go soon
Fe
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 3:
Loss Functions
and Optimization
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 3  1
April 11, 2017
Administrative
Assignment 1 is released:
http:/cs231n.github.io/assignments2017/assignment1/
Due Thursday April 20, 11:59pm on Canvas
(Exten
Convolutional Neural Networks for Visual Recognition
CS 231N

Winter 2014
Lecture 5:
Convolutional Neural Networks
FeiFei Li & Justin Johnson & Serena Yeung
Lecture 5  1
April 18, 2017
Administrative
Assignment 1 due Thursday April 20, 11:59pm on Canvas
Assignment 2 will be released Thursday
FeiFei Li & Justin Johnson & Sere