Self-supervised depth characterization method based on comparative learning

The invention discloses a self-supervised depth characterization method based on comparative learning. The method comprises the following steps: S1, obtaining a to-be-trained data set V; s2, training N pieces of data in batch at a time, and performing data enhancement on each piece of data v; s3, transmitting the data v and the enhanced data vs into a restricted Boltzmann machine for forward propagation to obtain data h and data hs corresponding to the restricted Boltzmann machine in a hidden layer; s4, performing comparative learning to improve the similarity between positive samples; s5, back propagation is carried out, and reconstruction data v'of the data v in a visual layer are obtained; s6, minimizing the distance between the data v'and the data v; s7, repeating the steps S2-S6 until the reconstruction loss function converges, and realizing the training of the restricted Boltzmann machine; and S8: using the restricted Boltzmann machine trained in S7 to carry out depth representation learning on the data