Tiny images representation and nearest neighbor classifier

Implementation

All training and test images were saved as 16x16 representations. Each test image was classified using a K nearest neighbors classifier for the training tiny images (each pixel was used as a feature). 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	0.020					Highway	Bedroom	Coast	Highway
Store	0.020					InsideCity	Coast	Forest	Forest
Bedroom	0.090					LivingRoom	Suburb	Coast	Coast
LivingRoom	0.050					Suburb	Kitchen	TallBuilding	Highway
Office	0.060					InsideCity	Bedroom	TallBuilding	Forest
Industrial	0.030					Office	InsideCity	Highway	Highway
Suburb	0.240					Coast	Bedroom	Street	Office
InsideCity	0.070					Mountain	OpenCountry	Highway	Coast
TallBuilding	0.090					Highway	Coast	Forest	Coast
Street	0.470					Suburb	TallBuilding	InsideCity	OpenCountry
Highway	0.800					OpenCountry	Suburb	OpenCountry	Coast
OpenCountry	0.300					Mountain	Street	Coast	Coast
Coast	0.340					Industrial	Bedroom	OpenCountry	OpenCountry
Mountain	0.130					LivingRoom	Forest	Highway	OpenCountry
Forest	0.380					Office	InsideCity	Coast	Highway
Category name	Accuracy	Sample training images		Sample true positives		False positives with true label		False negatives with wrong predicted label