eecs 442 – computer vision segmentation & clustering
TRANSCRIPT
![Page 1: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/1.jpg)
EECS 442 – Computer vision
Segmentation & Clustering
Reading: Chapters 14 [FP]
• Segmentation in human vision• K-mean clustering• Mean-shift• Graph-cut
Some slides of this lectures are courtesy of prof F. Li, profS. Lazebnik, and various other lecturers
![Page 2: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/2.jpg)
![Page 3: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/3.jpg)
Segmentation
• Compact representation for image data in terms of a set of components
• Components share “common” visual properties
• Properties can be defined at different level of abstractions
![Page 4: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/4.jpg)
General ideas
• Tokens
– whatever we need to group (pixels, points, surface elements, etc., etc.)
• Bottom up segmentation
– tokens belong together because they are locally coherent
• Top down segmentation
– tokens belong together because they lie on the same object
> These two are not mutually exclusive
![Page 5: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/5.jpg)
What is Segmentation?
• Clustering image elements that “belong together”
– Partitioning• Divide into regions/sequences with coherent
internal properties
– Grouping• Identify sets of coherent tokens in image
Slide credit: Christopher Rasmussen
![Page 6: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/6.jpg)
Why do these tokens belong together?
What is Segmentation?
![Page 7: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/7.jpg)
Basic ideas of grouping in human vision
• Figure-ground discrimination
• Gestalt properties
![Page 8: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/8.jpg)
– Grouping can be seen in terms of allocating some elements to a figure, some to ground
– Can be based on local bottom-up cues or high level recognition
Figure-ground discrimination
![Page 9: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/9.jpg)
Figure-ground discrimination
![Page 10: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/10.jpg)
– A series of factors affect whether elements should be grouped together
Gestalt properties
![Page 11: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/11.jpg)
Gestalt properties
![Page 12: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/12.jpg)
Gestalt properties
![Page 13: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/13.jpg)
Gestalt properties
Grouping by occlusions
![Page 14: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/14.jpg)
Gestalt properties
Grouping by invisible completions
![Page 15: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/15.jpg)
Emergence
![Page 16: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/16.jpg)
Segmentation in computer vision
J. Malik, S. Belongie, T. Leung and J. Shi. "Contour and Texture Analysis for Image Segmentation". IJCV 43(1),7-27,2001.
![Page 17: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/17.jpg)
Object Recognition as Machine Translation, Duygulu, Barnard, de Freitas, Forsyth, ECCV02
Segmentation in computer vision
![Page 18: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/18.jpg)
Segmentation as clustering
Cluster together tokens that share similar visual characteristics
• K-mean • Mean-shift• Graph-cut
![Page 19: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/19.jpg)
Feature Space
• Every token is identified by a set of salient visual characteristics. For example: – Position – Color– Texture– Motion vector– Size, orientation (if token is larger than a pixel)
Slide credit: Christopher Rasmussen
![Page 20: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/20.jpg)
Source: K. Grauman
Feature Space
![Page 21: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/21.jpg)
Feature space: each token is represented by a point
r
b
g
![Page 22: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/22.jpg)
r
b
g
Token similarity is thus measured by distance between points (“feature vectors”) in feature space
q
p
![Page 23: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/23.jpg)
r
b
g
Cluster together tokens with high similarity
![Page 24: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/24.jpg)
![Page 25: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/25.jpg)
![Page 26: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/26.jpg)
![Page 27: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/27.jpg)
![Page 28: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/28.jpg)
![Page 29: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/29.jpg)
![Page 30: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/30.jpg)
![Page 31: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/31.jpg)
• Agglomerative clustering– Add token to cluster if token is similar enough
to element of clusters– Repeat
• Divisive clustering– Split cluster into subclusters if along best
boundary– Boundary separates subclusters based on
similarity – Repeat
![Page 32: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/32.jpg)
Hierachical structure of clusters (Dendrograms)
• rand('seed',12); • X = rand(100,2); • Y = pdist(X,‘euclidean'); • Z = linkage(Y,‘single'); • [H, T] = dendrogram(Z);
![Page 33: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/33.jpg)
K-Means Clustering
• Initialization: Given K categories, N points in feature space. Pick K points randomly; these are initial cluster centers (means) m1, …, mK. Repeat the following:
1. Assign each of the N points, xj , to clusters by nearest mi
2. Re-compute mean mi of each cluster from its member points
3. If no mean has changed more than some ε, stop
Slide credit: Christopher Rasmussen
![Page 34: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/34.jpg)
Example: 3-means Clustering
Source: wikipedia
![Page 35: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/35.jpg)
• Effectively carries out gradient descent to minimize:
![Page 36: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/36.jpg)
![Page 37: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/37.jpg)
![Page 38: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/38.jpg)
![Page 39: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/39.jpg)
Image Clusters on intensity Clusters on color
K-means clustering using intensity alone and color alone
![Page 40: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/40.jpg)
K-Means pros and cons
• Pros– Simple and fast– Converges to a local minimum of the error function
• Cons– Need to pick K– Sensitive to initialization– Only finds “spherical”
clusters– Sensitive to outliers
![Page 41: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/41.jpg)
• An advanced and versatile technique for clustering-based segmentation
D. Comaniciu and P. Meer, Mean Shift: A Robust Approach toward Feature Space Analysis, PAMI 2002.
Mean shift segmentation
![Page 42: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/42.jpg)
• The mean shift algorithm seeks a mode or local maximum of density of a given distribution– Choose a search window (width and location)– Compute the mean of the data in the search window– Center the search window at the new mean location – Repeat until convergence
Mean shift segmentation
![Page 43: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/43.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Slide by Y. Ukrainitz & B. Sarel
![Page 44: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/44.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Slide by Y. Ukrainitz & B. Sarel
![Page 45: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/45.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 46: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/46.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 47: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/47.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Slide by Y. Ukrainitz & B. Sarel
![Page 48: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/48.jpg)
Region ofinterest
Center ofmass
Mean Shiftvector
Slide by Y. Ukrainitz & B. Sarel
![Page 49: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/49.jpg)
Region ofinterest
Center ofmass
Slide by Y. Ukrainitz & B. Sarel
![Page 50: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/50.jpg)
Computing The Mean Shift
Simple Mean Shift procedure:• Compute mean shift vector
•Translate the Kernel window by m(x)
2
1
2
1
( )
ni
ii
ni
i
gh
gh
=
=
⎡ ⎤⎛ ⎞⎢ ⎥⎜ ⎟⎜ ⎟⎢ ⎥⎝ ⎠= −⎢ ⎥⎛ ⎞⎢ ⎥⎜ ⎟⎢ ⎥⎜ ⎟
⎝ ⎠⎣ ⎦
∑
∑
x - xx
m x xx - x
g( ) ( )k′= −x x
![Page 51: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/51.jpg)
Real Modality Analysis
• Tessellate the space with windows
•Merge windows that end up near the same “peak” or model
![Page 52: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/52.jpg)
• Attraction basin: the region for which all trajectories lead to the same mode
• Cluster: all data points in the attraction basin of a mode
Slide by Y. Ukrainitz & B. Sarel
Attraction basin
![Page 53: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/53.jpg)
Attraction basin
Slide by Y. Ukrainitz & B. Sarel
![Page 54: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/54.jpg)
• Find features (color, gradients, texture, etc)• Initialize windows at individual pixel locations• Perform mean shift for each window until convergence• Merge windows that end up near the same “peak” or mode
Segmentation by Mean Shift
![Page 55: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/55.jpg)
http://www.caip.rutgers.edu/~comanici/MSPAMI/msPamiResults.html
Mean shift segmentation results
![Page 56: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/56.jpg)
![Page 57: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/57.jpg)
Mean shift pros and cons
• Pros– Does not assume spherical clusters– Just a single parameter (window size) – Finds variable number of modes– Robust to outliers
• Cons– Output depends on window size– Computationally expensive– Does not scale well with dimension of feature
space
![Page 58: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/58.jpg)
Graph-based segmentation
• Represent features and their relationships using a graph
• Cut the graph to get subgraphs with strong interior links and weaker exterior links
![Page 59: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/59.jpg)
Images as graphs
• Node for every pixel
• Edge between every pair of pixels
• Each edge is weighted by the affinity or similarity of the two nodes
wiji
j
Source: S. Seitz
![Page 60: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/60.jpg)
Measuring Affinity
Intensity
Color
Distance
aff x, y( )= exp − 12σ i
2⎛ ⎝
⎞ ⎠ I x( )− I y( ) 2( )⎧
⎨ ⎩
⎫ ⎬ ⎭
aff x, y( )= exp − 12σ d
2⎛ ⎝
⎞ ⎠ x − y 2( )⎧
⎨ ⎩
⎫ ⎬ ⎭
aff x, y( )= exp − 12σ t
2⎛ ⎝
⎞ ⎠ c x( )− c y( ) 2( )⎧
⎨ ⎩
⎫ ⎬ ⎭
![Page 61: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/61.jpg)
Segmentation by graph partitioning
• Break Graph into sub-graphs– Break links (cutting) that have low affinity
• similar pixels should be in the same sub-graphs• dissimilar pixels should be in different sub-graphs
A B C
Source: S. Seitz
wiji
j
• Sub-graphs represents different image segments
![Page 62: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/62.jpg)
Graph cut
• CUT: Set of edges whose removal makes a graph disconnected
• Cost of a cut: sum of weights of cut edges• A graph cut gives us a segmentation
– What is a “good” graph cut and how do we find one?
Source: S. Seitz
![Page 63: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/63.jpg)
Graph cut
• CUT: Set of edges whose removal makes a graph disconnected
• Cost of a cut: sum of weights of cut edges• A graph cut gives us a segmentation
– What is a “good” graph cut and how do we find one?
A B C
![Page 64: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/64.jpg)
Minimum cut
• We can do segmentation by finding the minimum cut in a graph– Efficient algorithms exist for doing this
• Drawback: minimum cut tends to cut off very small, isolated components
Ideal Cut
Cuts with lesser weightthan the ideal cut
* Slide from Khurram Hassan-Shafique CAP5415 Computer Vision 2003
![Page 65: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/65.jpg)
• IDEA: normalizing the cut by component size• The normalized cut cost is:
• The exact solution is NP-hard but an approximation can be computed by solving a generalized eigenvalue problem
assoc(A, V) = sum of weights of all edges in V that touch A
Normalized cut
),(),(
),(),(
VBassocBAcut
VAassocBAcut
+
J. Shi and J. Malik. Normalized cuts and image segmentation. PAMI 2000
![Page 66: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/66.jpg)
• Pros– Generic framework, can be used with many
different features and affinity formulations
• Cons– High storage requirement and time complexity– Bias towards partitioning into equal segments
* Slide from Khurram Hassan-Shafique CAP5415 Computer Vision 2003
Normalized cuts: Pro and con
![Page 67: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/67.jpg)
Figure from “Image and video segmentation: the normalised cut framework”, by Shi and Malik, copyright IEEE, 1998
Normalized cuts: Results
![Page 68: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/68.jpg)
F igure from “Normalized cuts and image segmentation,” Shi and Malik, copyright IEEE, 2000
Normalized cuts: Results
![Page 69: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/69.jpg)
J. Malik, S. Belongie, T. Leung and J. Shi. "Contour and Texture Analysis for Image Segmentation". IJCV 43(1),7-27,2001.
Contour and Texture Analysis for Image Segmentation
![Page 70: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/70.jpg)
Efficient Graph-Based Image Segmentation Pedro F. Felzenszwalb and Daniel P. HuttenlocherInternational Journal of Computer Vision, Volume 59, Number 2, September 2004
Efficient Graph-Based Image Segmentation
superpixel
![Page 71: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/71.jpg)
Integrating top-down and bottom-up segmentation
Z.W. Tu, X.R. Chen, A.L. Yuille, and S.C. Zhu. Image parsing: unifying segmentation, detection and recognition. IJCV 63(2), 113-140, 2005.
![Page 72: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/72.jpg)
EECS 442 – Computer vision
Object Recognition
![Page 73: EECS 442 – Computer vision Segmentation & Clustering](https://reader030.vdocuments.us/reader030/viewer/2022020119/5867803f1a28ab6d408bcb09/html5/thumbnails/73.jpg)