1. Human action recognition using spatio-temporal features Nikhil Sawant (2007MCS2899) Guide : Dr. K.K. Biswas
2. Human activity recognition Higher resolution Longer Time Scale Courtesy : Y. Ke, Fathi and Mori, Bobick and Davis, Schuldt et al, Leibe et al, Vaswani et al. Pose Estimation Action Recognition Action Classification Tracking Activity Recognition
23. Classification Example taken from Antonio Torralba @MIT Weak learners from the family of lines h => p(error) = 0.5 it is at chance Each data point has a class label: w t =1 and a weight: + 1 ( ) -1 ( ) y t =
24. Classification Example This one seems to be the best This is a ‘ weak classifier ’: It performs slightly better than chance. Each data point has a class label: w t =1 and a weight: + 1 ( ) -1 ( ) y t =
25. Classification Example We set a new problem for which the previous weak classifier performs at chance again Each data point has a class label: w t w t exp{-y t H t } We update the weights: + 1 ( ) - 1 ( ) y t =
26. Classification Example We set a new problem for which the previous weak classifier performs at chance again Each data point has a class label: w t w t exp{-y t H t } We update the weights: + 1 ( ) - 1 ( ) y t =
27. Classification Example We set a new problem for which the previous weak classifier performs at chance again Each data point has a class label: w t w t exp{-y t H t } We update the weights: + 1 ( ) - 1 ( ) y t =
28. Classification Example We set a new problem for which the previous weak classifier performs at chance again Each data point has a class label: w t w t exp{-y t H t } We update the weights: + 1 ( ) - 1 ( ) y t =
29. Classification Example The strong (non- linear) classifier is built as the combination of all the weak (linear) classifiers. f 1 f 2 f 3 f 4