Bag of SIFT representation and nearest neighbor classifier

Implementation

Features were extracted using a bag of SIFT representation. First we built a vocabulary by taking a subset of sift features of a sampling from the training images, then did k means samplings on this list of features to make K (I used 400 to balance time and accuracy) centroids of features. However, the actual features for each image are a histogram of samplings of sift features for that image binned into the nearest neighbor vocabulary word (centroid). Each test image was classified using a K nearest neighbors classifier for the bag of sifts features. Accuracy was optimized with K=4.

Results

Category name	Accuracy	Sample training images		Sample true positives		False positives with true label		False negatives with wrong predicted label
Kitchen	0.390					Industrial	Industrial	Bedroom	Office
Store	0.510					TallBuilding	TallBuilding	LivingRoom	Industrial
Bedroom	0.320					OpenCountry	Kitchen	LivingRoom	Office
LivingRoom	0.340					Bedroom	Store	Office	Office
Office	0.820					Kitchen	Industrial	Kitchen	Kitchen
Industrial	0.410					TallBuilding	Store	Kitchen	Store
Suburb	0.930					OpenCountry	Mountain	LivingRoom	Bedroom
InsideCity	0.450					Street	Suburb	Industrial	Street
TallBuilding	0.440					InsideCity	LivingRoom	OpenCountry	Street
Street	0.560					Bedroom	InsideCity	Mountain	Suburb
Highway	0.800					Coast	OpenCountry	Street	Suburb
OpenCountry	0.510					Street	Bedroom	Highway	Forest
Coast	0.560					Highway	OpenCountry	Highway	OpenCountry
Mountain	0.520					OpenCountry	Coast	Forest	Suburb
Forest	0.940					Industrial	OpenCountry	Suburb	Suburb
Category name	Accuracy	Sample training images		Sample true positives		False positives with true label		False negatives with wrong predicted label