Elemantra: An End-to-End Framework Empowered with Edge AI to Tackle Human-Elephant Conflict

The cohabitation of elephants and humans has evolved into a human-elephant conflict (HEC) due to the increasing loss of historical elephant habitations. As HEC is a substantial threat to both species, advanced sensing methods are utilized to develop HEC prevention frameworks which still lack unification. Here, we propose a collaborative framework for HEC prevention consisting of three main modules: (1) a joint infrared-seismic sensing strategy to facilitate elephant detection at the edge using a distributed and light-weight deep learning module, (2) an on-site repelling system with time-varying acoustic and light deterrents, and (3) a mesh network for edge device communication. Our end-to-end framework is designed to operate with no human intervention through a set of novel and systematic decision-making pipelines and algorithms. The preliminary results of each module confirm that the proposed framework is effective in achieving their individual, but complementary, goals and thereby promising a fully automated collaborative system for HEC prevention. We publicly release our annotated datasets and codes at https://github.com/nuwansribandara/elemantra to support the development of HEC prevention systems for the benefit of the environment.