Active Basis + Texture Model

What's new?

2. Add more figures for Exp 2.

Experiment 1 -- Adaptive Background Subtraction

Buddha Data AUC 0.887 --> 0.891

Head Shoulder Data AUC 0.977 --> 0.983

Experiment 2 -- Masked Texture Statistics

2.1 AUC

Head Shoulder Data AUC 0.977 --> 0.985

2.2 Visualization --- Buddha

Budda Data, mask

Budda Data, background distribution and Gaussian fitting

Budda Data, foreground distribution and Gaussian fitting

2.3 Visualization --- HS

HS, mask

HS Data, background distribution and Gaussian fitting

HS Data, foreground distribution and Gaussian fitting

Experiment 3 -- Masked Texture Statistics + Adaptive Background Subtraction

Buddha Data AUC 0.887 --> 0.906

Head Shoulder Data AUC 0.977 --> 0.986

Experiment 4 -- Meshed Texture Statistics

4.1 AUC

Buddha Data AUC 0.887 --> 0.918

Head Shoulder Data AUC 0.977 --> 0.974

4.2 Visualization --- Buddha

Buddha --- Denaulay trianglization, bottom, pathes of Mesh 1 and 2 in positive, negative, training, testing examples

Buddha --- Background Histogram of top 3 meshes, Gaussian or exponential fitting

Buddha --- Foreground Histogram of top 3 meshes, Gaussian or exponential fitting

Buddha --- Coding Gain

4.3 Visualization --- Head-Shoulder

HS --- Denaulay trianglization, bottom, pathes of Mesh 1 and 2 in positive, negative, training, testing examples

HS --- Background Histogram of top 3 meshes, Gaussian or exponential fitting

HS --- Foreground Histogram of top 3 meshes, Gaussian or exponential fitting

HS --- Coding Gain