Skip to content

Commit fca5970

Browse files
committed
Updates slides.
1 parent e28d9bb commit fca5970

2 files changed

Lines changed: 275 additions & 11 deletions

File tree

slides/slides.pdf

2.37 MB
Binary file not shown.

slides/slides.tex

Lines changed: 275 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44
\usepackage[T1]{fontenc}
55

66
\usepackage{tikz}
7-
\usetikzlibrary{positioning}
7+
\usetikzlibrary{matrix, positioning}
88
\usetikzlibrary{decorations.pathreplacing}
99
\usepackage{etoolbox}
1010
\usepackage{listofitems} % for \readlist to create arrays
@@ -96,9 +96,9 @@
9696
\frame{\tableofcontents}
9797

9898

99-
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
100-
% ----------- Intro al DL -------------------
101-
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
99+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
100+
% ----------- Intro al DL --------------------
101+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
102102
\section{Intro al aprendizaje profundo}
103103

104104
% ----------- Motivación 01 ------------------
@@ -1418,13 +1418,6 @@
14181418
\texttt{history = model.fit(..., callbacks=[callback])}
14191419
\end{frame}
14201420

1421-
% Mover/eliminar conforme se agregan los contenidos
1422-
\subsection{Regularización}
1423-
\section{Visión computacional profunda}
1424-
\section{Modelado profundo de secuencias}
1425-
\section{Modelado generativo profundo}
1426-
\section{Panorama actual y futuro}
1427-
14281421
% ----------- Lecturas recomendadas 03 ------
14291422
\begin{frame}{Lecturas recomendadas}{Intro al aprendizaje profundo}
14301423
\begin{itemize}
@@ -1433,6 +1426,277 @@
14331426
\item \colorbox{blue!10}{\href{https://neptune.ai/blog/vanishing-and-exploding-gradients-debugging-monitoring-fixing}{Vanishing and Exploding Gradients in Neural Network Models:}} Debugging, Monitoring, and Fixing
14341427
\end{itemize}
14351428
\end{frame}
1429+
1430+
1431+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1432+
% ----------- Computer Vision ----------------
1433+
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1434+
\section{Visión computacional profunda}
1435+
1436+
% ----------- Intro a imágenes 01 ------------
1437+
\subsection{Introducción a imágenes}
1438+
\begin{frame}{¿Qué es una imagen?}{Visión computacional profunda}
1439+
\begin{figure}
1440+
\centering
1441+
\includegraphics[width=0.75\textwidth]{img-deer}
1442+
\end{figure}
1443+
\end{frame}
1444+
1445+
% ----------- Intro a imágenes 02 ------------
1446+
\begin{frame}{¿Qué es una imagen?}{Visión computacional profunda}
1447+
\begin{figure}
1448+
\centering
1449+
\includegraphics[width=0.75\textwidth]{img-lincoln}
1450+
\let\thefootnote\relax\footnote{{\tiny “Tutorial 1: Image Filtering.” AI Stanford. \url{https://ai.stanford.edu/~syyeung/cvweb/tutorial1.html}}}
1451+
\end{figure}
1452+
\end{frame}
1453+
1454+
% ----------- Intro a imágenes 03 ------------
1455+
\begin{frame}{¿Qué es una imagen?}{Visión computacional profunda}
1456+
\begin{figure}
1457+
\centering
1458+
\includegraphics[width=0.75\textwidth]{img-lena}
1459+
\let\thefootnote\relax\footnote{{\tiny “Tutorial 1: Image Filtering.” AI Stanford. \url{https://ai.stanford.edu/~syyeung/cvweb/tutorial1.html}}}
1460+
\end{figure}
1461+
\end{frame}
1462+
1463+
% ----------- Intro a imágenes 04 ------------
1464+
\begin{frame}{¿Qué es una imagen?}{Visión computacional profunda}
1465+
\begin{itemize}
1466+
\item Una imagen es un arreglo de pixeles, la cual peude tener 1 o más canales de color. Usualmente:
1467+
\begin{itemize}
1468+
\item 1 canal de color $\rightarrow$ Escala de grises
1469+
\item 3 canales de color $\rightarrow$ Escala RGB
1470+
\item 4 canales de color $\rightarrow$ Escala RGBA
1471+
\end{itemize}
1472+
\item Un pixel puede ser visto como un objeto 5-dimensional $(x, y, r, g, b)$.
1473+
\end{itemize}
1474+
\end{frame}
1475+
1476+
% ----------- Biología humana 01 -------------
1477+
\begin{frame}{La biología humana}{Visión computacional profunda}
1478+
\begin{figure}
1479+
\centering
1480+
\includegraphics[width=0.8\textwidth]{img-encroma-01}
1481+
\let\thefootnote\relax\footnote{{\tiny “How EnChroma Color Blind Glasses Work.” EnChroma. \url{http://enchroma.com/technology/}}}
1482+
\end{figure}
1483+
\end{frame}
1484+
1485+
% ----------- Biología humana 02 -------------
1486+
\begin{frame}{La biología humana}{Visión computacional profunda}
1487+
\begin{figure}
1488+
\centering
1489+
\includegraphics[width=0.8\textwidth]{img-encroma-02}
1490+
\let\thefootnote\relax\footnote{{\tiny “How EnChroma Color Blind Glasses Work.” EnChroma. \url{http://enchroma.com/technology/}}}
1491+
\end{figure}
1492+
\end{frame}
1493+
1494+
% ----------- Espacios de color --------------
1495+
\subsection{Espacios de color}
1496+
\begin{frame}{Espacios de color}{Visión computacional profunda}
1497+
\begin{figure}
1498+
\centering
1499+
\includegraphics[width=0.9\textwidth]{img-color-space}
1500+
\end{figure}
1501+
\end{frame}
1502+
1503+
% ----------- Convoluciones 01 ---------------
1504+
\subsection{Convoluciones \& Pooling}
1505+
\begin{frame}{¿Qué es una convolución?}{Visión computacional profunda}
1506+
\begin{figure}
1507+
\centering
1508+
\includegraphics[width=0.9\textwidth]{1d-conv}
1509+
\end{figure}
1510+
\end{frame}
1511+
1512+
% ----------- Ejercicio 05 ------------------
1513+
\begin{frame}{Ejercicio}{Intro al aprendizaje profundo}
1514+
\begin{center}
1515+
{\Large \textbf{Ejercicio: Introducción a imágenes}}
1516+
\end{center}
1517+
\begin{figure}
1518+
\centering
1519+
\includegraphics[width=0.6\textwidth]{img-lena}
1520+
\end{figure}
1521+
\end{frame}
1522+
1523+
% ----------- Convoluciones 02 ---------------
1524+
\begin{frame}{¿Qué es una convolución?}{Visión computacional profunda}
1525+
\begin{center}
1526+
\begin{tikzpicture}[
1527+
2d-arr/.style={matrix of nodes, row sep=-\pgflinewidth, column sep=-\pgflinewidth, nodes={draw}}, ampersand replacement=\&
1528+
]
1529+
1530+
\matrix (mtr) [2d-arr] {
1531+
0 \& 1 \& 1 \& |[fill=orange!30]| 1 \& |[fill=orange!30]| 0 \& |[fill=orange!30]| 0 \& 0\\
1532+
0 \& 0 \& 1 \& |[fill=orange!30]| 1 \& |[fill=orange!30]| 1 \& |[fill=orange!30]| 0 \& 0\\
1533+
0 \& 0 \& 0 \& |[fill=orange!30]| 1 \& |[fill=orange!30]| 1 \& |[fill=orange!30]| 1 \& 0\\
1534+
0 \& 0 \& 0 \& 1 \& 1 \& 0 \& 0\\
1535+
0 \& 0 \& 1 \& 1 \& 0 \& 0 \& 0\\
1536+
0 \& 1 \& 1 \& 0 \& 0 \& 0 \& 0\\
1537+
1 \& 1 \& 0 \& 0 \& 0 \& 0 \& 0\\
1538+
};
1539+
1540+
\node[below=of mtr-5-4] {$\mathbf I$};
1541+
1542+
\node[right=0.2em of mtr] (str) {$*$};
1543+
1544+
\matrix (K) [2d-arr, right=0.2em of str, nodes={draw, fill=teal!30}] {
1545+
1 \& 0 \& 1 \\
1546+
0 \& 1 \& 0 \\
1547+
1 \& 0 \& 1 \\
1548+
};
1549+
\node[below=of K-3-2] {$\mathbf K$};
1550+
1551+
\node[right=0.2em of K] (eq) {$=$};
1552+
1553+
\matrix (ret) [2d-arr, right=0.2em of eq] {
1554+
1 \& 4 \& 3 \& |[fill=blue!80!black!30]| 4 \& 1\\
1555+
1 \& 2 \& 4 \& 3 \& 3\\
1556+
1 \& 2 \& 3 \& 4 \& 1\\
1557+
1 \& 3 \& 3 \& 1 \& 1\\
1558+
3 \& 3 \& 1 \& 1 \& 0\\
1559+
};
1560+
\node[below=of ret-4-3] {$\mathbf{I * K}$};
1561+
1562+
\draw[dashed, teal] (mtr-1-6.north east) -- (K-1-1.north west);
1563+
\draw[dashed, teal] (mtr-3-6.south east) -- (K-3-1.south west);
1564+
1565+
\draw[dashed, blue!80!black] (K-1-3.north east) -- (ret-1-4.north west);
1566+
\draw[dashed, blue!80!black] (K-3-3.south east) -- (ret-1-4.south west);
1567+
1568+
\foreach \i in {1,2,3} {
1569+
\foreach \j in {4,5,6} {
1570+
\node[font=\tiny, scale=0.6, shift={(-1.2ex,-2ex)}] at (mtr-\i-\j) {$\times \pgfmathparse{int(mod(\i+\j,2))}\pgfmathresult$};
1571+
}
1572+
}
1573+
1574+
\end{tikzpicture}
1575+
\end{center}
1576+
\end{frame}
1577+
1578+
% ----------- Convoluciones 03 ---------------
1579+
\begin{frame}{¿Qué es una convolución?}{Visión computacional profunda}
1580+
\begin{figure}
1581+
\centering
1582+
\includegraphics[width=0.9\textwidth]{filter}
1583+
\end{figure}
1584+
\let\thefootnote\relax\footnote{{\tiny “Image Kernels.” Victor Powell. \url{https://setosa.io/ev/image-kernels/}}}
1585+
\end{frame}
1586+
1587+
% ----------- Pooling 01 ---------------------
1588+
\begin{frame}{Pooling}{Visión computacional profunda}
1589+
\begin{figure}
1590+
\centering
1591+
\includegraphics[width=0.9\textwidth]{pooling}
1592+
\end{figure}
1593+
\end{frame}
1594+
1595+
% ----------- Ejercicio 06 ------------------
1596+
\begin{frame}{Ejercicio}{Intro al aprendizaje profundo}
1597+
\begin{center}
1598+
{\Large \textbf{Ejercicio: Convoluciones \& Pooling}}
1599+
\end{center}
1600+
\begin{figure}
1601+
\centering
1602+
\includegraphics[width=0.6\textwidth]{filter}
1603+
\end{figure}
1604+
\end{frame}
1605+
1606+
% ----------- CNNs 01 ------------------------
1607+
\subsection{Redes neuronales convolucionales}
1608+
\begin{frame}{Redes neuronales convolucionales}{Visión computacional profunda}
1609+
\begin{figure}
1610+
\centering
1611+
\includegraphics[width=0.9\textwidth]{lecun}
1612+
\caption{Yann LeCun}
1613+
\end{figure}
1614+
\end{frame}
1615+
1616+
% ----------- CNNs 02 ------------------------
1617+
\begin{frame}{Redes neuronales convolucionales}{Visión computacional profunda}
1618+
\begin{figure}
1619+
\centering
1620+
\includegraphics[width=0.9\textwidth]{cnns-01}
1621+
\end{figure}
1622+
\end{frame}
1623+
1624+
% ----------- CNNs 03 ------------------------
1625+
\begin{frame}{Redes neuronales convolucionales}{Visión computacional profunda}
1626+
\begin{figure}
1627+
\centering
1628+
\includegraphics[width=0.9\textwidth]{cnns-02}
1629+
\end{figure}
1630+
\end{frame}
1631+
1632+
% ----------- CNNs 04 ------------------------
1633+
\begin{frame}{Redes neuronales convolucionales}{Visión computacional profunda}
1634+
\begin{figure}
1635+
\centering
1636+
\includegraphics[width=0.9\textwidth]{cnns-03}
1637+
\end{figure}
1638+
\end{frame}
1639+
1640+
% ----------- CNNs 05 ------------------------
1641+
\subsection{Clasificadores de imágenes (LeNet, VGG16, etc.)}
1642+
\begin{frame}{LeNet-5}{Visión computacional profunda}
1643+
\begin{figure}
1644+
\centering
1645+
\includegraphics[width=0.9\textwidth]{classifiers-01}
1646+
\end{figure}
1647+
\end{frame}
1648+
1649+
% ----------- CNNs 06 ------------------------
1650+
\begin{frame}{VGG16}{Visión computacional profunda}
1651+
\begin{figure}
1652+
\centering
1653+
\includegraphics[width=0.9\textwidth]{classifiers-02}
1654+
\end{figure}
1655+
\end{frame}
1656+
1657+
% ----------- CNNs 07 ------------------------
1658+
\begin{frame}{Resnet50}{Visión computacional profunda}
1659+
\begin{figure}
1660+
\centering
1661+
\includegraphics[width=0.9\textwidth]{classifiers-03}
1662+
\end{figure}
1663+
\end{frame}
1664+
1665+
% ----------- CNNs 08 ------------------------
1666+
\begin{frame}{GoogLeNet}{Visión computacional profunda}
1667+
\begin{figure}
1668+
\centering
1669+
\includegraphics[width=0.9\textwidth]{classifiers-04}
1670+
\end{figure}
1671+
\end{frame}
1672+
1673+
% ----------- Ejercicio 07 ------------------
1674+
\begin{frame}{Ejercicio}{Intro al aprendizaje profundo}
1675+
\begin{center}
1676+
{\Large \textbf{Ejercicio: Redes neuronales convolucionales}}
1677+
\end{center}
1678+
\begin{figure}
1679+
\centering
1680+
\includegraphics[width=0.6\textwidth]{classifiers-01}
1681+
\end{figure}
1682+
\end{frame}
1683+
1684+
1685+
% Mover/eliminar conforme se agregan los contenidos
1686+
\section{Modelado profundo de secuencias}
1687+
\section{Modelado generativo profundo}
1688+
\section{Panorama actual y futuro}
1689+
1690+
% ----------- Lecturas recomendadas 04 ------
1691+
\begin{frame}{Lecturas recomendadas}{Visión computacional}
1692+
\begin{itemize}
1693+
\item Tutorial 1: \colorbox{blue!10}{\href{https://ai.stanford.edu/~syyeung/cvweb/tutorial1.html}{Image Filtering}}
1694+
\item \colorbox{blue!10}{\href{https://setosa.io/ev/image-kernels/}{Image Kernels Explained Visually}} by Victor Powell
1695+
\item \colorbox{blue!10}{\href{https://gudgud96.github.io/2020/11/25/param-pooling/}{Parameterized Pooling Layers}} by Hao Hao Tan
1696+
\item TensorFlow Tutorials: \colorbox{blue!10}{\href{https://www.tensorflow.org/tutorials/images}{Vision}}
1697+
1698+
\end{itemize}
1699+
\end{frame}
14361700

14371701

14381702
\end{document}

0 commit comments

Comments
 (0)