Self-Supervised Convolutional Subspace Clustering Network with the Block Diagonal Regularizer

The practical visual data do not necessarily lie in linear subspaces, so deep convolutional subspace clustering network is proposed to segment the into multiple categories accurately. original contains stacked encoder module, decoder module and self-expression module. We firstly alter i.e., add a new k-block diagonal regularizer weights of It means that $$\ell _1$$ or _2$$ abandoned. directly pursue block matrix, introducing this will make learned representation matrix conform with better. Secondly, we spectral network, which result used supervise learning matrix. This structured introduced further refines Experimental results on three challenging datasets have demonstrated based method achieves better effect over state-of-the-arts.