SVCL Motion Classification

Library Walk ... Scene Description - [html] [pdf]
Congestion Level Experiments Oblique Angle #2 Still				Experiment Accuracy
				Martin NN = 82% Nearest-neighbor classifier using Martin distance. State KL NN = 82% Nearest-neighbor classifier using state KL divergence State KL SVM = 83% SVM classifier using state KL kernel Image KL NN = 62% Nearest-neighbor classifier using image KL divergence Image KL SVM = 87% SVM classifier using image KL kernel
Description In this experiment, we attempt to discriminate between three congestion levels of pedsetrian traffic using still clips at an oblique angle on a different day as the first (below). Location: 4th floor of Geisel Library, UC San Diego.	Clips: High Flow divx mov mp4 divx 1 2 mov (h.264) 1 2 mp4 (mpeg4) 1 2	Clips: Medium Flow divx mov mp4 divx 1 2 mov (h.264) 1 2 mp4 (mpeg4) 1 2	Clips: Low Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Oblique Angle #1 Still				Experiment Accuracy
				Martin NN = 76% Nearest-neighbor classifier using Martin distance. State KL NN = 76% Nearest-neighbor classifier using state KL divergence State KL SVM = 88% SVM classifier using state KL kernel Image KL NN = 60% Nearest-neighbor classifier using image KL divergence Image KL SVM = 86% SVM classifier using image KL kernel
Description In this experiment, we attempt to discriminate between three congestion levels of pedsetrian traffic using still clips at an oblique angle. Location: 4th floor of Geisel Library, UC San Diego.	Clips: High Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1	Clips: Medium Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1	Clips: Low Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Oblique Angle #3 Still			Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
	Clips: High Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1	Clips: Medium Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Front Angle Still				Experiment Accuracy
				Martin NN = 64% Nearest-neighbor classifier using Martin distance. State KL NN = 64% Nearest-neighbor classifier using state KL divergence State KL SVM = 58% SVM classifier using state KL kernel Image KL NN = 62% Nearest-neighbor classifier using image KL divergence Image KL SVM = 70% SVM classifier using image KL kernel
Description In this experiment, we attempt to discriminate between three different levels of pedestrian traffic flow using still clips at a front angle.	Clips: High Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1	Clips: Medium Flow divx mov mp4 divx 1 2 mov (h.264) 1 2 mp4 (mpeg4) 1 2	Clips: Low Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Front Angle #2 Still			Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description In this experiment, we attempt to discriminate between three different levels of pedestrian traffic flow using still clips at a front angle.	Clips: High Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1	Clips: Medium Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Side Angle Still			Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description In this experiment, we attempt to discriminate between different levels of pedestrian traffic flow using still clips at a side angle.	Clips: Low Flow divx mov mp4 divx 1 2 3 4 5 mov (h.264) 1 2 3 4 5 mp4 (mpeg4) 1 2 3 4 5	Clips: No Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Side Angle #2 Still		Need other classes to discriminate.	Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description In this experiment, we attempt to discriminate between different levels of pedestrian traffic flow using still clips at a side angle.	Clips: Low Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Front Angle Eye Level Walking		Need other classes to discriminate.	Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description No experiment setup yet.	Clips: High Flow divx mov mp4 divx 1 2 3 mov (h.264) 1 2 3 mp4 (mpeg4) 1 2 3

Front Angle Above Head Level Walking		Need other classes to discriminate.	Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description No experiment setup yet.	Clips: High Flow divx mov mp4 divx 1 2 3 mov (h.264) 1 2 3 mp4 (mpeg4) 1 2 3

Side Angle Pan		Need other classes to discriminate.	Need other classes to discriminate.	Experiment Accuracy
				No Experimental Results
Description No experiment setup yet.	Clips: Flow divx mov mp4 divx 1 mov (h.264) 1 mp4 (mpeg4) 1

Oblique Angle #3 Still

Oblique Angle #3
Still