Quick Start

To get started quickly with Glimpse, use the Glab API. As an example, we perform object detection on the AnimalDB dataset used by Serre et al [1], using an HMAX-like model. (This was run on a 2.33 GHz Intel Core2 Quad CPU.)

>>> from glimpse.glab import *
>>> Verbose()
>>> image_dir = "/path/to/AnimalDB"
>>> SetCorpus(image_dir)
>>> ImprintS2Prototypes(100)
Imprinting 100 prototypes
  done: 18.6740250587 s
>>> RunSvm()
Train SVM on 600 images
  and testing on 600 images
  done: 0.471097946167 s
Time to compute feature vectors: 322.276679039 s
Accuracy is 0.917 on training set, and 0.792 on test set.

References

[1]	Serre, T., Oliva, A. & Poggio, T., 2007. A feedforward architecture accounts for rapid categorization. Proceedings of the National Academy of Sciences, 104(15), p.6424-6429.