Efficient Aspect Object Models Using Pre-trained Convolutional Neural Networks 


Humanoids 2015

Object Dataset




Download the data set here

The data set file contains all segmented images, type labels, and features. Features were compressed and stored using the python library  Hickle. The props.p file stores a python dictionary and contains the type, object, and aspect ids for each image in the data set. Images were obtained using an ASUS XTion camera.