Identifying corresponding SAR-Optical patches using Inception-inspired Pseudo-Siamese CNN
Dhruv Patidar, Nitin S. Sharma · IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium · 2022
Even for humans, at times interpreting a SAR image can prove to be quite challenging, let alone finding and identifying the corresponding optical patch. The problem of identifying corresponding SAR-optical patches finds many applications in remote sensing and computer vision. While there have been attempts to solve the problem with the use of convolutional neural networks like Siamese networks, the excellence of more recent advances can be used to solve it more effectively. This paper is an attempt towards the same. We use Inception-inspired convolutional modules in a Pseudo-Siamese network to make progress in solving the problem of identifying corresponding patches in SAR and Optical images.