Project 3: Object categorization with a “bag of
keypoints”
Description
•
Implement
a bag of ‘words’ approach described in “Visual
Categorization with Bags of Keypoints” G.Cruska, C. R. Dance, L.Fan,
J.Willamowski,C. Bray.
•
Test
it on 4 categories: airplanes, faces, cars side, and motorbikes.
•
Optional:
train the system to detect the above categories against general
background.
Results
The detection results should
be presented as a confusion matrix:
|
|
Airplanes |
Faces |
Cars |
Motorbikes |
|
Airplanes |
How
many of the planes are classified as planes |
How
many of the faces are classified as planes |
… |
… |
|
Faces |
How
many of the planes are classified as faces |
How
many of the faces are classified as faces |
… |
… |
|
Cars |
How
many of the planes are classified as cars |
How
many of the faces are classified as cars |
… |
… |
|
Motorbikes |
How
many of the planes are classified as motorbikes |
How
many of the faces are classified as motorbikes |
… |
… |
Obviously the best results
will produce diagonal confusion matrix.
Data sets
The system should be trained
and tested on the above 4 categories from
Caltech
101 objects data set. You can
download the background images (optional) from here.
SVM Package
K-Means
Use MATLAB function kmeans(…) for
the codebook construction.
Useful Links
Class lecture on object
detection
"Distinctive image features from
scale-invariant keypoints" by David Lowe
Status
Claimed by Ran and Elran