Refresh lab4_segmentation

lab01_git/questions.md

@@ -1,70 +1,88 @@
 # Warmup Questions
 
 1.  What is the clone url of this repository?
-    >   answer
+    >   https://github.com/CarlosAAcostaH/lab_vision.git
 
 2.  What is the output of the ``cal`` command?
 
-        multi
-        line
-        answer
+           February 2016
+           Su Mo Tu We Th Fr Sa
+           1  2  3  4  5  6
+           7  8  9 10 11 12 13
+           14 15 16 17 18 19 20
+           21 22 23 24 25 26 27
+           28 29
+
+
 
 # Homework Questions
 
 1.  What is the ``grep``command?
-    >   answer
+    >
+        "The grep utility searches any given input files, selecting lines that
+         match one or more patterns"
+                                                                       -man grep
 
 2.  What is a *makefile*?
-    >   answer
+    >   A *makefile* is a text file used to built or manange proyects with commands in C/C++
 
 4.  What does the ``-prune`` option of ``find`` do? Give an example
-    >   answer
+    >   
+            ``-prune`` causes ``find`` to not descend into the current file.
+            Ex.: find . -name .ingnore_Dir -prune -o -name '*.xyz' -print
+            It will find files with xyz extention but ignore the directory `.ignore_Dir`
 
 5.  Where is the ``grub.cfg``  file
-    >   answer
+    >   On ubuntu exist 3 ubications: /etc/grub.d, /usr/lib/grub and /usr/share/grub. On MacOS don't exist.
 
 6.  How many files with ``gnu`` in its name are in ``/usr/src``
-    >   answer
+    >   zero (find /usr/src/ -name gnu | wc -l)
 
 7.  How many files contain the word ``gpl`` inside in ``/usr/src``
-    >   answer
+    >   52 files (grep -lr "gpl" /usr/src/ | wc -l)
 
 8.  What does the ``cut`` command do?
-    >   answer
+    >   
+        "cut out selected portions of each line of a file"
+                                                    -man cut
 
 9.  What does the ``wget`` command do?
-    >   answer
+    >   Download files from the Web
 
 9.  What does the ``rsync`` command do?
-    >   answer
+    >   Copy files localy and to/from another host
 
 10.  What does the ``diff`` command do?
-    >   answer
+    >   
+        "Compare files line by line"
+                               -man diff
 
 10.  What does the ``tail`` command do?
-    >   answer
+    >  
+        "Print  the  last  10  lines of each FILE to standard output"
+                                                            -man tail
 
 10.  What does the ``tail -f`` command do?
-    >   answer
+    >   Show the new lines in the file in real time
 
 10.  What does the ``link`` command do?
-    >   answer
+    >   Creates ``linked files``
 
 11.  How many users exist in the course server?
-    >   answer
+    >   30 users
 
 12. What command will produce a table of Users and Shells sorted by shell (tip: using ``cut`` and ``sort``)
-    >   answer
+    >   cut -d':' -f1,7 /etc/passwd | sort | tr ':' '\t'
 
 13. What command will produce the number of users with shell ``/sbin/nologin`` (tip: using ``grep`` and ``wc``)
-    >   answer
+    >   grep '/sbin/nologin' /etc/passwd | cut -d':' -f1 | wc -l
 
 15. Create a script for finding duplicate images based on their content (tip: hash or checksum)
     You may look in the internet for ideas, but please indicate the source of any code you use
     Save this script as ``find_duplicates.sh`` in this directory and commit your changes to github
 
 16. What is the meaning of ``#! /bin/bash`` at the start of scripts?
-    >   answer
+    >   The first two characters ``#!`` mean that the file is executable. The rest of the line mean the path where the script is execute.
 
 17. How many unique images are in the ``sipi_images`` database?
     >   answer

lab3_hybrid/Homework

@@ -0,0 +1,8 @@
+#Imagenes Originales
+|[]
+|[]
+#Imagenes modificadas con GIMP(R)
+|[]
+|[]
+#Imagen Hibrida Generada
+|[]

lab3_hybrid/homework.md

@@ -0,0 +1,65 @@
+# Hybrid Images
+## Information about the place
+The images were taken in diferents times in a same place: a Campito de San Jose's Chapel (hereafer called Chapel). The chapel was built in 1922 and later, in 1998, it was restaurated and converted in library. The first chapel's blueprints was known in 1903 and the bulding was acquired by the Universidad de los Andes in 1962. Nowdays, the chapel is the U building and works as Arts and Desing's library. 
+For more information about the chapel history go to [this link](https://campusinfo.uniandes.edu.co/images/stories/campus/Historia/Bloques/bloque%20u%20-%20capilla.pdf).
+The chapel is locate intro the Universidad de los Andes Campus in the east of Bogota, on the hillside of Guadalupe. The cartesian coordinates of the chapel are latitude 4.601518 N and longitude 74.064162 W.
+<div class="video-container">
+<iframe src="https://www.google.com/maps/embed?pb=!1m18!1m12!1m3!1d819.4706569373759!2d-74.06471217084585!3d4.60170233657302!2m3!1f0!2f0!3f0!3m2!1i1024!2i768!4f13.1!3m3!1m2!1s0x0%3A0x0!2zNMKwMzYnMDYuMSJOIDc0wrAwMyc1MS4wIlc!5e1!3m2!1ses-419!2s!4v1456885017487" width="600" height="450" frameborder="0" style="border:0" allowfullscreen></iframe>
+</div>
+## Originals Images
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/ViejaCapillaOrg.jpg" height="200" align="right">
+The Old Chapel's photo belongs to Oficina de Administración Documental (OAD) of Univerisdad de los Andes, it was taken at 1969 by Gemán Téllez. The original photo can be at this [link](https://www.facebook.com/155723537956531/photos/a.240801929448691.1073741833.155723537956531/240801939448690/?type=3&theater).
+The photo show was modificated (watermark) to protect the author rigths.
+For more information go to [OAD's facebook page](https://www.facebook.com/Oficina-de-Administración-Documental-155723537956531/) or [OAD's page](oad.uniandes.edu.co).
+
+The features of the old picture (as focal length and expotition time) do not know. The photo was obtained to the [OAD's facebook page](https://www.facebook.com/Oficina-de-Administración-Documental-155723537956531/)
+
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/NuevaCapillaOrg.jpg" height="200" align="left">
+The New Chapel's photo was taken by me, at 10:16 hrs (UTC-05:00), the day February 24th, 2016. The photo was taken with a cellphone's camera with the next features:
+- Dimention: 2988 × 5312 px
+- Color Space: RGB
+- Color Profile: sRGB IEC61966-2.1
+- Focal length: 4.8
+- F number: 2.2
+- Exposition time: 1/1.156 seg
+
+## Processed images
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/ViejaCapilla.jpg" height="200" align="left">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/NuevaCapilla.jpg" height="200" align="right">
+The images was processed with the tool GIMP&reg; to crop and fix the perspective. 
+
+The images' color did not modificate.
+
+The original images did not have the same perspective and therefor the size of windows, door, corners, stairs and others details were not similar. The perspective correction was used to relate the main facade of the chapel on the two images.
+
+The perspective correction can be do with MATLAB&reg; with a 3x3 matrix to make a geometric transformation, but is easier in GIMP&reg; because the tool calculate the 3x3 matrix based on the move of points on the images.
+
+## Final hybrid image
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="320" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="240" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="160" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="120" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="80" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="60" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="40" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="30" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="20" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="15" align="top">
+<img src="https://github.com/CarlosAAcostaH/lab_vision/blob/master/lab3_hybrid/HibridaCapilla.jpg" width="10" align="top">
+
+## Commented Code
+The code was write in m-language and run in MacOs Matlab R2015b. The processed images was saved in a directory `fotos`, placed in the same path to the code. The old photo was called `NuevaCapilla.jpg` and the new photo was called `ViejaCapilla.jpg`. The code save a hybrid image created as a jpg image (`HibridaCapilla.jpg`) in the directory `fotos`.
+```matlab
+clear all; % Clear all variables
+close all; % Close all open figures
+I1=imread('./fotos/NuevaCapilla.jpg'); % Create I1 as a 2016 chapel photo after modificated with GIMP(R)
+I2=imread('./fotos/ViejaCapilla.jpg'); % Create I2 as a 1969 chapel photo after modificated with GIMP(R)
+a=40; % Set an a parameter of the filter to apply to I1
+b=40; % Set a b parameter of the filter to apply to I2
+F1=fspecial('gaussian',floor(2*a+1)*[1 1],a*0.5); % Create a low-pass filter F1 that will be apply to I1
+F2=fspecial('gaussian',floor(2*b+1)*[1 1],b*0.5); % Create a low-pass filter F1 that will be apply to I2
+G1=imfilter(I1,F1); % Apply the filter F1 to I1 and save as G1
+G2=imfilter(I2,F2); % Apply the filter F2 to I2 and save as G2
+imshow(G1+(I2-G2)); % Show the hybrid image
+imwrite(G1+(I2-G2),'./fotos/HibridaCapilla.jpg') % Save the hybrid image created in a jpg file
+```

lab3_hybrid/lab3_hybrid.m

@@ -0,0 +1,12 @@
+clear all; % Clear all variables
+close all; % Close all open figures
+I1=imread('./fotos/NuevaCapilla.jpg'); % Create I1 as a 2016 chapel photo after modificated with GIMP(R)
+I2=imread('./fotos/ViejaCapilla.jpg'); % Create I2 as a 1969 chapel photo after modificated with GIMP(R)
+a=40; % Set an a parameter of the filter to apply to I1
+b=40; % Set a b parameter of the filter to apply to I2
+F1=fspecial('gaussian',floor(2*a+1)*[1 1],a*0.5); % Create a low-pass filter F1 that will be apply to I1
+F2=fspecial('gaussian',floor(2*b+1)*[1 1],b*0.5); % Create a low-pass filter F1 that will be apply to I2
+G1=imfilter(I1,F1); % Apply the filter F1 to I1 and save as G1
+G2=imfilter(I2,F2); % Apply the filter F2 to I2 and save as G2
+imshow(G1+(I2-G2)); % Show the hybrid image
+imwrite(G1+(I2-G2),'./fotos/HibridaCapilla.jpg') % Save the hybrid image created in a jpg file

lab4_segmentation/segment_by_clustering.m

@@ -0,0 +1,116 @@
+function my_segmentation = segment_by_clustering(I, feature_space,...
+    method, nClusters,isDisplayed)
+%% Probe that I is a rgb image
+if ((size(I,3)~=3)||(numel(size(I))~=3))
+    error (['The 1� input argument (I) must be a rgb image \nthe size'...
+        ' must be mxnx3%s'],'');
+elseif ~isa(I,'uint8')
+    error (['The 1� input argument (I) must be a rgb image \nI must'...
+        ' be a uint8, not a %s'], class(I))
+end
+%% Assign a value to nCluster
+if nargin<4
+    nClusters=3;
+end
+%% Probe that nClusters is a numeric
+if ~isa(nClusters,'numeric')
+    error(['The 4� input argument (nClusters) must be a numeric \n'...
+        'nClusters must be a numeric, not a %s'], class(nClusters))
+end
+%% If it is not a watershed clustering, change the data type of I to double
+if ~strcmp(method,'watershed')
+    fsHandle=1;
+    if strcmp (feature_space,'lab')
+        I=rgb2lab(I);
+        fsHandle=2;
+    elseif strcmp (feature_space,'hsv')
+        I=rgb2hsv(I);
+        fsHandle=3;
+    elseif strcmp (feature_space,'rgb+xy')
+        I=double(I);
+        fsHandle=4;
+    elseif strcmp (feature_space,'lab+xy')
+        I=rgb2lab(I);
+        fsHandle=5;
+    elseif strcmp (feature_space,'hsv+xy')
+        I=rgb2hsv(I);
+    elseif ~strcmp (feature_space,'rgb')
+        warning ('The feature space is wrong, it will be rgb by defect');
+        I=double(I);
+    else
+        I=double(I);
+    end
+%% Normalize and divide the the 3 channels of color
+    if (fsHandle==1||fsHandle==4)
+        ch1=I(:,:,1)/255;
+        ch2=I(:,:,2)/255;
+        ch3=I(:,:,3)/255;
+    elseif (fsHandle==2||fsHandle==5)
+        ch1=I(:,:,1)/100;
+        ch2=I(:,:,2);
+        ch3=I(:,:,3);
+        ch2=(ch2-min(ch2(:)))/(max(ch2(:))-min(ch2(:)));
+        ch3=(ch3-min(ch3(:)))/(max(ch3(:))-min(ch3(:)));
+    else
+        ch1=I(:,:,1);
+        ch2=I(:,:,2);
+        ch3=I(:,:,3);
+    end
+%% Create a feature space (fSpace)
+    if fsHandle<4
+        fSpace = [ch1(:) ch2(:) ch3(:)];
+    else
+        [x,y] = meshgrid(0:(size(I,2)-1),0:(size(I,1)-1));
+        r=1;
+        fSpace = [ch1(:) ch2(:) ch3(:) r*x(:)/(max(x(:))) r*y(:)/(max(y(:)))];
+    end
+%% If it is a watershed clustering convert I to gray-scale image    
+else
+    I=rgb2gray(I);
+end
+%% GMM clustering
+if strcmp(method,'gmm')
+    options = statset('MaxIter',10000);
+    gmfit = fitgmdist(fSpace,nClusters,'CovarianceType','full',...
+            'SharedCovariance',false,'Options',options);
+    idx = cluster(gmfit,fSpace);
+%% Watershed clustering
+elseif strcmp(method,'watershed')
+    hy = fspecial('sobel');
+    hx = hy';
+    Iy = imfilter(double(I), hy, 'replicate');
+    Ix = imfilter(double(I), hx, 'replicate');
+    grad = sqrt(Ix.^2 + Iy.^2); % Gradient of I
+    h=0;
+    while 1 % Search h to have a correct nCluster
+        marker=imextendedmin(I,h);
+        L=bwlabeln(marker);
+        if max(L(:))<=nClusters
+            break
+        end
+        h=h+1;
+    end
+    new_grad=imimposemin(grad,marker);
+    ws=watershed(new_grad);
+    idx=ws(:);
+%% KMEANS
+else
+    if ~strcmp(method,'kmeans')
+        warning ('The method is wrong, it will be kmeans by defect');
+    end
+    idx = kmeans(fSpace,nClusters,'MaxIter',10000);
+end
+%% Reshape the output
+my_segmentation=reshape(idx,[size(I,1) size(I,2)]);
+%% Show the clustering
+if nargin<5
+    isDisplayed=false;
+end
+if ~isa(isDisplayed,'logical')
+    error(['The 5� input argument (isDisplayed) must be a boolean \n'...
+    'isDisplayed must be a logical, not a %s'], class(isDisplayed))
+end
+if isDisplayed
+    imshow(label2rgb(my_segmentation,'jet','w','shuffle'));
+end
+end

lab5_bsds/homework.md

@@ -0,0 +1,29 @@
+1. Métodos de segmentación: 
+
+ Descripción básica de los métodos. Alternativas para el espacio de segmentación. ¿Qué métodos escogieron? ¿Por qué?
+ ### Método K-means
+ Es el método de segmentación mas simple y segmenta en k clusters donde k es un parametro. Su funcionamiento depende de los centroides iniciales. El algoritmo de este método es:
+  1. Definir k centroides en el el espacio de segmentación
+  2. Realizar la partición del espacio de segmentación con un diagrama de Voronoi con los k centroides
+  3. Recalcular el centroide de las k celdas como el promedio de los puntos pertenecientes a la celda de Voronoi
+  4. Si los centroides cambian, vuelva a paso ii; de lo contrario termino.
+ Al finalizar el algoritmo, el punto pertenece al cluster correspondiente al centroide mas cercano.
+
+### Método GMM
+Al igual que en el método de K-means, se recibe como parametro el numero de clusters deseados. En este caso no se asigna la pertenencia de un punto a un cluster durante la ejecución del método sino que se le asigna la probabilidad de pertenecer a un cluster y al finalizar se le asigna la etiqueta correspondiente al cluster más probable.
+
+###Método Watershed
+En este método se convierte la imagen con un sólo descriptor a un relieve del cual se encuentra las líneas de división de aguas. Para este método se convierte en escala de grises y el relieve es la magnitud del gradiente de la imagen en escaladegrises
+
+2. Metodología de pruebas: 
+
+ Descripción de la base de datos: ¿Cuántas imágenes son? ¿De qué tamaño? ¿De qué tipo ? ¿Cómo estaba organizada? ¿Cuál era la verdad terreno?; ¿Qué era lo que se medía en el benchmark? También rectificar que hubieran usado adecuadamente los conjuntos de train y test; especialmente que no hubieran iterado sobre "test".
+3. Presentación resultados: 
+
+ ¿Corrieron el benchmark con éxito? ¿Presentan imágenes segmentadas para ilustrar los procedimientos? ¿Describen con palabras los aspectos importantes que se ven en las imágenes y las gráficas?  
+4. Discusión: 
+
+ ¿Cómo se comparan los algoritmos? ¿A qué creen que se deben las diferencias? ¿Cuáles son las implicaciones prácticas de los resultados? ¿Qué método recomiendan? ¿Cuáles son las limitaciones de los métodos?
+5. Mejoras: 
+
+ ¿Cómo se podrían obtener mejores resultados en la segmentación? (acá pueden ser muy creativos)