WEEK 1-2 ALEJANDRO TORROELLA

CONVERTING AN IMAGE FROM RGB TO HSV AND DISPLAYING THE SEPARATE CHANNELS

Hue: Value: Saturation:

ROBERTS AND SOBEL EDGE DETECTORS ON BOTH RGB AND HSV CHANNELS

LAPLACIAN OF GAUSSIAN AND SEPARABILITY

Sigma = 1.2, Kernel Size = 9x9

CANNY EDGE DETECTOR

Sigma = 1.2, Kernel Size = 9x9, Threshold = 0.05

IMPROVING LEGIBILITY

HARRIS CORNER DETECTOR

SIFT DENSE SAMPLING SCHEME

SVM TRAINING AND TESTING

Used two different kernels: Chi Squared and Histogram Intersections Histogram Intersection gave better results than Chi Squared. Results were still poor, but that’s probably because of low density SIFT descriptors and the way they were implemented. Increasing the density of the SIFT descriptors significantly improved the accuracy of classification. Accuracy_768 = [ histo, chi2 ] = [ , ] Accuracy_3072 = [ histo, chi2 ] = [ , ]

OPTICAL FLOW

=

THANK YOU FIN.