03.11.2014 Views

Lecture 15 - Stanford Vision Lab

Lecture 15 - Stanford Vision Lab

Lecture 15 - Stanford Vision Lab

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Foreground model<br />

Generative probabilistic model<br />

Clutter model<br />

Gaussian shape pdf Prob. of detection Uniform shape pdf # detections<br />

0.8 0.75<br />

Assumptions: (a) Clutter independent of foreground detections<br />

(b) Clutter detections independent of each other<br />

Example<br />

1. Object Part Positions<br />

2. Part Absence 3a. N false detect<br />

0.9<br />

p Poisson (N 1 | 1 )<br />

p Poisson (N 2 | 2 )<br />

p Poisson (N 3 | 3 )<br />

3b. Position f. detect<br />

N 1<br />

N 2<br />

N 3<br />

Fei-Fei Li<br />

<strong>Lecture</strong> <strong>15</strong> -<br />

64<br />

14‐Nov‐11

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!