Nick Hunter Motivation Current Study Experiment Setup Video Clustering System Database Development Shot Transition Detection Shot Transition Detection Threshold Motion Haar‐Cascade Classification Haar‐like features, Neural Network and AdaBoost Development Clustering PCA‐Eigenface Recognition Eigenfaces Neighboring Compression Recognition Conclusions/Future works Recent Studies Showed an video/image to a patient Scanned their brain for activity From the brain scan they used the activity to guess and reconstruct the video/image shown Reconstruction : was of natural images based on both the structure and semantic content of the images simultaneously Haar‐Cascade Classifiers will be used to find and collect faces of many characters, also help develop character specific classifiers. Shot Transition Detection will help with: when to break tracking and when to start a new tracking set, which helps maintain where the object is within a frame‐set. K‐means theory to group the images of the characters to send to a database. Principal Component Analysis (PCA) from the collection of faces this learning algorithm will use eigenfaces to create a recognition database. prior and images were: current Converted to 8‐bit images Gaussian Blurred | ‐ | Compared to a white 255 level image of the same size. From this ratio disparity there suggest a level of motion activity in the movie. Frames Below Threshold 8000 6000 4000 , , (| 2000 50 100 ‐ , , 150 |/ 720*480*0.256) > Threshold 200 Threshold x 720 480 0.256 MIT+CMU test set containing 130 images and 507 faces <‐Euclidean distance is not a multivariate effect size Mahalanobis distances DFFS and DIFS allow for probabilistic interpretations ‐> (a) Aligned face. (b) Eigenspace reconstruction (85 bytes) (c) JPEG reconstruction (530 bytes) More understanding of the Robustness of each Algorithm Further work is needed on Video motion to: Better detect shot transitions Know if clustering characters is viable Database development Currently has only character segregation features Need to develop other semantic types to search for in the Brain [1] T. Naselaris, R. J. Prenger, K. N. Kay, M. Oliver, and J. L. Gallant, “Bayesian Reconstruction of Natural Images from Human Brain Activity,” Neuron, vol. 63, no. 6, pp. 902–915, Sep. 2009. [2] M. J. Chadwick, D. Hassabis, N. Weiskopf, and E. A. Maguire, “Decoding Individual Episodic Memory Traces in the Human Hippocampus,” Current Biology, vol. 20, no. 6, pp. 544–547, Mar. 2010. [3] S. Audet, http://code.google.com/p/javacv/, Sept. 2012. [4] T. Burton, Alice in Wonderland. Walt Disney Pictures, 2010. [5] Z. Rasheed and M. Shah, “Scene detection in Hollywood movies and TV shows,” in 2003 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2003. Proceedings, 2003, vol. 2, p. II– 343–8 vol.2. [6] R. Lienhart and J. Maydt, “An extended set of Haar‐like features for rapid object detection,” in 2002 International Conference on Image Processing. 2002. Proceedings, 2002, vol. 1, p. I–900– I–903 vol.1. [7] O. Arandjelovic and A. Zisserman, “Automatic Face Recognition for Film Character Retrieval in Feature‐ Length Films,” in Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05) ‐ Volume 1 ‐ Volume 01, Washington, DC, USA, 2005, pp. 860–867. [8] Qing Chen, N. D. Georganas, and E. M. Petriu, “Real‐time Vision‐based Hand Gesture Recognition Using Haar‐ like Features,” in IEEE Instrumentation and Measurement Technology Conference Proceedings, 2007. IMTC 2007, 2007, pp. 1–6. [9] R. Szeliski, Computer Vision: Algorithms and Applications, 1st ed. Springer, 2010. [10] Rainer Lienhart and Jochen Maydt. An Extended Set of Haar‐like Features for Rapid Object Detection. IEEE ICIP 2002, Vol. 1, pp. 900‐903, Sep. 2002. http://www.lienhart.de/ICIP2002.pdf [11] P. Viola and M. Jones, “Rapid Object Detection Using a Boosted Cascade of Simple Features,” 2001, pp. 511– 518. [12] M. Kirby and L. Sirovich, “Application of the Karhunen‐Loeve procedure for the characterization of human faces,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 12, no. 1, pp. 103 –108, Jan. 1990. http://en.wikipedia.org/wiki/Fmri http://en.wikipedia.org/wiki/Bayes%27_theorem Visual Image Reconstruction from Human Brain Activity using a Combination of Multiscale Local Image Decoders Blood‐oxygen‐level dependent (BOLD) is the MRI contrast of blood deoxyhemoglobin PROS noninvasively record brain signals resolution can be as good as 1mm. Localized recordings of signals of the brain fMRI is widely used and standard data‐analysis approaches have been developed which allow researchers to compare results across labs. fMRI produces compelling images of brain "activation". http://en.wikipedia.org/wiki/Fmri