Building Recognition on Subregion’s Multiscale Gist Feature Extraction and Corresponding Columns Information Based Dimensionality Reduction
Figure 8
The recognition accuracies of LPP and CCIG-LPP versus the dimensions when three and four images were randomly selected for each category for training. In (a) and (c) three respective images of each category were randomly selected for training. In (b) and (d) four respective images of each category were randomly selected for training. In (a) and (b) SM-gist was used to extract gist features. In (c) and (d) we use Siagian and Itti’s feature extraction model to extract gist features.
