Micro Segmentation Security: Securing IT Through Macro-segmentation
Segmentation
description
Transcript of Segmentation
![Page 1: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/1.jpg)
Segmentation
CS4670 / 5670: Computer VisionNoah Snavely
From Sandlot Science
![Page 2: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/2.jpg)
Consumer face application: Apple iPhoto (Picasa, Facebook, etc.)
http://www.apple.com/ilife/iphoto/
![Page 3: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/3.jpg)
Consumer application: Apple iPhoto
Can be trained to recognize pets!
http://www.maclife.com/article/news/iphotos_faces_recognizes_cats
![Page 4: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/4.jpg)
Consumer application: Apple iPhoto
Things iPhoto thinks are faces
![Page 5: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/5.jpg)
Funny Nikon ads"The Nikon S60 detects up to 12 faces."
![Page 6: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/6.jpg)
Lambda Labs
![Page 7: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/7.jpg)
Image segmentation
![Page 8: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/8.jpg)
The goals of segmentation
• Group together similar-looking pixels for efficiency of further processing• “Bottom-up” process
X. Ren and J. Malik. Learning a classification model for segmentation. ICCV 2003.
“superpixels”
![Page 9: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/9.jpg)
The goals of segmentation
• Separate image into coherent “objects”• “Bottom-up” or “top-down” process?
Berkeley segmentation database:http://www.eecs.berkeley.edu/Research/Projects/CS/vision/grouping/segbench/
image human segmentation
![Page 10: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/10.jpg)
The Gestalt school
• Elements in a collection can have properties that result from relationships • “The whole is greater than the sum of its parts”
Illusory contours occlusion
familiar configuration
http://en.wikipedia.org/wiki/Gestalt_psychology
![Page 11: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/11.jpg)
Emergence
http://en.wikipedia.org/wiki/Gestalt_psychology
![Page 12: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/12.jpg)
Gestalt factors
• These factors make intuitive sense, but are very difficult to translate into algorithms
![Page 13: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/13.jpg)
Segmentation as clustering
Source: K. Grauman
![Page 14: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/14.jpg)
Image Intensity-based clusters Color-based clusters
Segmentation as clustering
• K-means clustering based on intensity or color is essentially vector quantization of the image attributes• Clusters don’t have to be spatially coherent
![Page 15: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/15.jpg)
Segmentation as clustering
Source: K. Grauman
![Page 16: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/16.jpg)
Segmentation as clustering
• Clustering based on (r,g,b,x,y) values enforces more spatial coherence
![Page 17: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/17.jpg)
K-Means for segmentation
• Pros• Very simple method• Converges to a local minimum of the error function
• Cons• Memory-intensive• Need to pick K• Sensitive to initialization• Sensitive to outliers• Only finds “spherical”
clusters
![Page 18: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/18.jpg)
http://www.caip.rutgers.edu/~comanici/MSPAMI/msPamiResults.html
Mean shift clustering and segmentation
• 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.
![Page 19: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/19.jpg)
• The mean shift algorithm seeks modes or local maxima of density in the feature space
Mean shift algorithm
imageFeature space
(L*u*v* color values)
![Page 20: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/20.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 21: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/21.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 22: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/22.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 23: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/23.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 24: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/24.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 25: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/25.jpg)
Searchwindow
Center ofmass
Mean Shiftvector
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 26: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/26.jpg)
Searchwindow
Center ofmass
Mean shift
Slide by Y. Ukrainitz & B. Sarel
![Page 27: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/27.jpg)
• Cluster: all data points in the attraction basin of a mode
• Attraction basin: the region for which all trajectories lead to the same mode
Mean shift clustering
Slide by Y. Ukrainitz & B. Sarel
![Page 28: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/28.jpg)
• Find features (color, gradients, texture, etc)• Initialize windows at individual feature points• Perform mean shift for each window until convergence• Merge windows that end up near the same “peak” or mode
Mean shift clustering/segmentation
![Page 29: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/29.jpg)
http://www.caip.rutgers.edu/~comanici/MSPAMI/msPamiResults.html
Mean shift segmentation results
![Page 30: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/30.jpg)
More results
![Page 31: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/31.jpg)
More results
![Page 32: Segmentation](https://reader035.fdocuments.us/reader035/viewer/2022062807/5681519e550346895dbfd098/html5/thumbnails/32.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