1 paper
Muhammad Shakeel, Katsutoshi Itoyama, Kenji Nishida +1
We describe a novel metric-based learning approach that introduces a multimodal framework and uses deep audio and geophone encoders in siamese configuration to design an adaptable…