svms for (x) recognition (from moghaddam / yang’s “gender classification with svms”) brian...
TRANSCRIPT
![Page 1: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/1.jpg)
SVMs for (x) Recognition
(From Moghaddam / Yang’s “Gender Classification with SVMs”)
Brian Whitman
![Page 2: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/2.jpg)
“Commodity Intelligence” ‘Wow factor’ important
Collaborative filtering ‘Simple’ tasks sometimes the most useful An SVM embedded evaluator… Cameras with ‘common sense’
![Page 3: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/3.jpg)
Why SVM for feature detection? Quick evaluation model Machines (SVs) are easily stored and small
![Page 4: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/4.jpg)
Experiment: Gender ID Using MITFaces dataset
~7500 faces with varying genders, races, ages, expressions, ‘extras’
All aligned 160x160 with left eye at 80,80 Face content is usually only 80x40
![Page 5: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/5.jpg)
MITFaces examples
![Page 6: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/6.jpg)
Representation? Simple pixel values Why?
![Page 7: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/7.jpg)
Sample size Maintain ‘ground rule’ of ML
Dimensions < Examples*2 At 3200 dims (80x40), this is hard Training parameters (maximum
lagrangians, kernel width) help We use 80x40 and 40x20 in our examples
![Page 8: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/8.jpg)
Training stage Choose 3200 random adult faces for
training and 3200 random faces for testing Extract 80x40 ‘face window’ from each
face and treat the 3200 doubles (0..1) as a training example
Train SVM on pixel values of the train set (dual p4 xeon linux 2ghz -- 30 minutes)
![Page 9: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/9.jpg)
Testing Stage Take the other 3200 face vectors and
present them to the learned SVM If class > 0, male, < 0, female. Confidence: some linear combination of #
of support vectors and magnitude of result Had no problem doing this at 10hz on a
PIII800 with tons running
![Page 10: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/10.jpg)
In-class face gender results 80x40; C=100, aux=100
93% of faces classified correctly 95% male 90% female
40x20; C=100, aux=10 97%
98% male 95% female
![Page 11: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/11.jpg)
Next step: Realtime Media Lab is where webcams go to die Webcam at 160x120, ‘face region’ to
80x40, downsampled to 40x20. Webcam gets frames at 10hz, we greyscale
it and present it to the previously trained SVM
Results… mixed
![Page 12: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/12.jpg)
Realtime examples (If the demo crashes)
![Page 13: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/13.jpg)
‘Creepybot’ With better control over alignment Monitors Windows clipboard Same architecture as the Creepycam
![Page 14: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/14.jpg)
Creepybot Examples (If the demo crashes)
![Page 15: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/15.jpg)
Other parameters MITFaces has a great data label set Train an SVM for appearance of each
descriptor: Race Age Gender Expression Moustache
![Page 16: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/16.jpg)
Per-class results (40x20, etc…) “Adult or not”
Overall: 94% (Not adult: 403/516) (78%) (Adult): 2605/2684) (97%)
![Page 17: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/17.jpg)
Per-class results… “Smiling or not”
Overall: 88% (Not smiling: 1354/1520) (89%) (Smiling: 1450 / 1672) (87%)
![Page 18: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/18.jpg)
Per-class results “Serious or not”
Overall: 88% (Not serious: 1517/1712) (89%) (Serious: 1311/1484) (88%)
![Page 19: SVMs for (x) Recognition (From Moghaddam / Yang’s “Gender Classification with SVMs”) Brian Whitman](https://reader035.vdocument.in/reader035/viewer/2022070411/56649f445503460f94c64c83/html5/thumbnails/19.jpg)
Could we do better? Representation is lacking But results are surprisingly good For realtime, need auto-alignment / rescaling, or a
better representation Could this lead to an invasion of cheap intelligent
cameras, each with tacky switches for feature detection and marketing?