How Can Selection of Biologically Inspired Features Improve the Performance of a Robust Object Recognition Model?
Figure 6
The performance which was obtained with different number of training examples for our model and the HMAX model.
(A), The performances for one positive training image (all other image sets, negative training, positive test and negative test consist of 50 images), in most data sets the performance for all three cases are almost equal. (B, C), The performance for 3 and 6 training examples respectively. (D, E, F), Performance of the models for 15, 30, and 40 training images respectively. White bars show the performance of the HMAX model with P randomly extracted patches and the red bars illustrate the proposed model performance, cyan bars show the performance of the HMAX model for 1000 patches.