Spectra Clustering and Image Segmentation

Inspector custeing, you interpret your image as a grap so every pixle is a note. So spectral clustering then identifies clusters within that grave. And spectra clustering, an cluster is defined as something which is easy to cut out. On the deep learning side, do we apply that same technique to transform our dat into a space where the edges are now realizable by the traditional norl networks? I think there's nothing like that. Explicitly, you have just these convolutional layes, which are good for translation in variancs. There's also some work on making newer networks forome votational invariant on that ifyou witate an