Implicit neural representation for change detection

08 Mar 2023 in Publications

Unsupervised urban sprawling change detection with implicit neural representation with a two step method. Shape reconstruction for support correction followed by simple difference of the two shapes leading to the detected change. The findings of the paper are: 1) this method beats state of the art and is close to the supervised schemes, 2) an agreement loss term allows to reconciliate and utilises both timestamped point clouds raw information.

Short summary

Unsupervised urban sprawling change detection with implicit neural representation with a two step method. Shape reconstruction for support correction followed by simple difference of the two shapes leading to the detected change. The findings of the paper are: 1) this method beats state of the art and is close to the supervised schemes, 2) an agreement loss term allows to reconciliate and utilises both timestamped point clouds raw information.