Results without Segmentation

Posted: May 2, 2016 at 6:27 pm

The following images show the results of training a SOM using the colour values of the original pano pixels and bypassing the segmentation processes. The result is that performance is improved (a 5px cell is possible), and the images are quite interesting, although the white sky dominates the image, where it was previously ignored as it was seen as a single giant segment and ignored.

SOMResults_noSegmentation_SOMScale5_h1352_ns400_i1000000-scaler8SOMResults_noSegmentation_SOMScale5_h1352_ns200_i1000000-scaler8

The difference between the preceding images is the max neighbourhood size, which is 400 on top and 200 on bottom. The smaller neighbourhood size is quite interesting and I particularly like how the area at the horizon breaks down. I (hopefully) shot the final panorama today (~250 RAW images with bracketing) and am working on stitching them together now, more on that later.