Fundamentals of Machine Learning

Chenyi Chen

What’s learning? • Example problem: face recognition

Prof. K

Prof. F

Prof. P

Prof. V

Chenyi

• Training data: a collection of images and labels (names)

Who is this guy?

• Evaluation criterion: correct labeling of new images

What’s learning? • Example problem: scene classification

road

road

sea

mountain

city

• a few labeled training images

What’s the label of this image?

• goal to label yet unseen image

Why learning? • The world is very complicated • We don’t know the exact model/mechanism between input and output • Find an approximate (usually simplified) model between input and output through learning • Principles of learning are “universal” – society (e.g., scientific community) – animal (e.g., human) – machine

A Taste of Machine Learning

True label

Estimated label We want to minimize the difference between them!

Artificial Neural Network

Fundamentals of Computer Vision

Chenyi Chen

What is Computer Vision? • Input: images • Output: information about the world

What is Computer Vision? Example: • What is in this image? • Who is in this image? • Where are they? • What are they doing?

What is Computer Vision? Other questions: • What camera settings were used? • Which pixels go with which objects? • What is the scene description in 3D?

Camera Projection

Camera Projection

Camera Projection

Dimensionality Reduction Machine (3D to 2D) 3D world

2D image

Point of observation

What have we lost? Slide by A. Efros Figures © Stephen E. Palmer, 2002

A Tale of Two Coordinate Systems v

COP w

y

u

Camera

Two important coordinate systems: 1. World coordinate system z 2. Camera coordinate system

x

o “The World”

Geometric Transformations

What is the geometric relationship between these two images?

?

Image alignment

Why don’t these image line up exactly?

What is the geometric relationship between these two images?

Very important for creating mosaics!

2D image transformations

These transformations are a nested set of groups • Closed under composition and inverse is a member

Projective Transformations / Homographies

Called a homography (or planar perspective map)

Image warping with homographies

image plane in front

image plane below black area where no pixel maps to

Homographies

A Quick Application: Lane Detection

Lane Detection

Lane Detection

Stereo Vision

A Taste of Stereo Vision

Visual Odometry, Structure-from-Motion, 3D Street Scene Reconstruction

KITTI Datasets • • • • • •

Stereo images Grayscale Color Rectified 1382*512 10 FPS

Visual Odometry • Visual odometry computes the trajectory of the vehicle only based on image sequences (LIBVISO2)

Depth Map • Disparity map is computed from grayscale stereo image pairs (LIBELAS) • Depth map can be derived from disparity map and camera model

Lane Detection • Projecting lane markers on the road (Caltech Lane Detector)

3D Street Scene Reconstruction

+ +

3D Street Scene Reconstruction • Dense reconstruction on run_70

Reconstruction with Non-Stereo Images/Structure-from-Motion • Triangulation: tracking a same point in three (or more) frames, its spatial position can be determined

Figure courtesy of Jianxiong Xiao

Sparse Reconstruction with Non-Stereo Images • Sparse reconstruction on run_70

Sparse Reconstruction with Non-Stereo Images • run_1

Sparse Reconstruction with Non-Stereo Images • run_9

Other Demos for Structure-from-Motion • https://www.youtube.com/watch?v=i7ierVkXY a8 • https://www.youtube.com/watch?v=vpTEobp YoTg

Other Demos for Structure-from-Motion

Other Demos for Structure-from-Motion

Deep Learning

The end of all the fundamentals

Vision Based Self-driving Car - Princeton University

The world is very complicated. • We don't know the exact model/mechanism between input and output. • Find an approximate (usually simplified) model between input and output through learning. • Principles of learning are “universal”. – society (e.g., scientific community). – animal (e.g., human). – machine ...
Missing:

6MB Sizes 0 Downloads 231 Views

Recommend Documents

New Immigrants.qxd - Princeton University
Westminster Register Office. Standing beside a framed photo ..... satellite television, cheap phone calls and the internet, people in developing countries are more ...

Painting with Triangles - Princeton Graphics - Princeton University
By con- trast, programs like Adobe Illustrator and Inkscape let a user paint ... effect. These arbitrary polygons are costly to render however, and. “smooth” effects are only created via many .... Next the outer stroke polygons are rendered in 50

Attenuation of Adaptation - Princeton University
strategy, it cannot capture the rapid initial reduction in error or the overcompensatory drift. Therefore, we modeled the strategy as a setpoint/reference signal that ...

New Immigrants.qxd - Princeton University Press
2. Immigrants: Your Country Needs Them grey-haired man in a bright red, fur-trimmed robe ..... ing for a few years in the Valley and set up companies that trade.

net neutrality - cs.Princeton - Princeton University
Jul 6, 2006 - of traffic when your browser needs to fetch a new page from a server. If a network provider is using ... hand, applications like online gaming or Internet telephony (VoIP), which rely on steady streaming of interactive .... The VPN user

New Immigrants.qxd - Princeton University Press
months while learning English; a forty-six-year-old Romanian dental technician who described ..... remote locations to complete their degree courses online. And.

Chapter 2 [PDF] - Princeton University Press
enables us to apply the tools used in the analysis of stationary models to study economies with sustained ...... of computer hardware and software. Thus we may ...

Prior Expectations Bias Sensory ... - Princeton University
In a separate analysis, we estimated BOLD amplitudes for each single trial, using the .... weights, we take our training data Bloc and regress those onto our hypo-.

Chapter 2 [PDF] - Princeton University Press
For more information send email to: ... These economists published two pathbreaking articles in the same year, 1956 (Solow, 1956;. Swan .... will study in Chapter 8) is that technology is free: it is publicly available as a nonexcludable, ... Definit

Prior Expectations Bias Sensory ... - Princeton University
segment in a 360° circle (Fig. 1A) using two buttons of an MR-compatible button box to rotate the line clockwise or anticlockwise. The initial di- rection of the line ...

Exporting and Organizational Change - Princeton University
Jul 18, 2017 - The computations in this paper were done at a secure data center .... of management (or L + 1 layers of employees, given that we call the ... of length z costs ¯wcz (c teachers per unit of knowledge at cost ¯w per teacher).

Exporting and Organizational Change - Princeton University
Jul 18, 2017 - We study the effect of exporting on the organization of production within firms. .... their technology (and so the marginal product of labor is higher) or .... Learning how to solve problems in an interval of knowledge .... We use conf

When Human Rights Pressure is ... - Princeton University
Sep 9, 2016 - have access to a good VPN to access foreign websites (Time 2013). On the other side, Chinese ... Times 2015), which was debated on social media. .... comparison to AU criticism 19, while AU criticism itself has no effect20.

Sources of Wage Inequality - Princeton University
Jan 14, 2013 - strong empirical support. Helpman et al. ... facts that support the mechanism of firm$ ..... An International Comparison, Chicago: University of ...

Vision-based hexagonal image processing based hexagonal ... - IJRIT
addresses and data of hexagonal pixels. As shown in Fig. 2, the spiral architecture is inspired from anatomical consideration of the primate's vision system.

pdf-12115\propaganda-and-the-cold-war-a-princeton-university ...
... problem loading more pages. Retrying... pdf-12115\propaganda-and-the-cold-war-a-princeton-university-symposium-from-brand-literary-licensing-llc.pdf.

Trade, Merchants, and the Lost Cities of the ... - Princeton University
Jun 27, 2017 - multiple ancient cities within their boundary. Using 2014 ..... The lower panel presents simple statistics (mean, minimum and maximum). 32 ...

z DFbz zz DF y − + - Princeton University
... and JA Taylor, Dept. of Biomedical Eng., Washington University, St. Louis, MO .... t r. Field One. Field Two. Field Four. -200. -150. -100. -50. 0. 50. 100. 150.

Empirical evidence on inflation expectations in ... - Princeton University
and time period, but with revised data, reduces the estimate on the activity ... share as the proxy for firm marginal cost) around the turn of the millennium was ...... He demonstrates in an empirical application that his method can give very ..... B

Life-Cycle Wage Growth Across Countries - Princeton University
May 22, 2016. Abstract ..... different time trends. Figure 1 plots experience-wage profiles ..... business-cycle effects that average to zero over the long run” (p.126).

VISION-BASED CONTROL FOR AUTONOMOUS ...
data, viz. the mean diameter of the citrus fruit, along with the target image size and the camera focal length to generate the 3D depth information. A controller.

VISION-BASED CONTROL FOR AUTONOMOUS ... - Semantic Scholar
invaluable guidance and support during the last semester of my research. ..... limits the application of teach by zooming visual servo controller to the artificial ... proposed an apple harvesting prototype robot— MAGALI, implementing a spherical.

Vision-based hexagonal image processing based hexagonal image ...
computer vision and pattern Recognition, Las Vegas, June 2006. [8] R.M. Mersereau, “The processing of Hexagonally Sampled Two-. Dimensional Signals,” Proceedings of the IEEE. 67: pp. 930 949, 1979. [9] X. He and W. Jia, “hexagonal structure for

Stereo Vision based Robot Navigation
stereo vision to guide a robot that could plan paths, construct maps and explore an indoor environment. ..... versity Press, ISBN: 0521540518, second edi-.