Segmentation Results
Figure 5: The link scheme used for segmentation. Neurons with identically sized receptive fields are arranged in layers. Neurons with partially overlapping receptive fields are connected by uni- or bidirectional synaptic links.
The dynamic linking scheme described in the previous page can be used with a specific link structure to analyze scale space (figure 5). Arranging all neurons in layers of fixed resolution, two types of links are introduced: unidirectional interlevel links, connecting coarser resolutions levels with the next finer levels, and bidirectional intralevel links, connecting neighboring neurons within a layer. Due to the local nature of the link scheme, only oscillators close in scale space can couple with each other, thus forcing synchronized activity to follow contour sheets.
Figure 6: Two segmentation results. In (a), the input to the network was the same is in figure 1. In b), a picture from a standard dataset (CIL-0001, S.A. Shaffer, Calibrated Imaging Lab, Carnegie Mellon University.) was used.
In order to perform large scale simulations a simplified locking scheme was used. Two units coupled by uni- or bidirectional links were marked as synchronized, if the difference in their firing rates was less than a prescribed threshold. In case of bidirectional links, the common frequency of the synchronized units was assumed to be the average of the individual modulation frequencies before synchronization. During linking, units with the smallest frequency difference were allowed to synchronize first. Units at the end of unidirectional links were just set to the modulation frequency of the driving units if the sync-threshold was not exceeded.
The results of two segmentation runs are depicted in figure 6.
The neural network consisted of a total of six resolution layers with the
link structure proposed in figure 5.
The only parameters of the model, the inter- and intralevel link thresholds,
were kept fixed for both simulation runs (all intralevel thresholds were
set at 
, except 
, and all interlevel thresholds at 
, except 
).
| Prev: Neural Implementation | ^ Up: Table of Contents ^ | Next: Acknowledgements |
Comments are welcome! © 1997 by Rolf Henkel - all rights reserved.