Inproceedings,

Multi-Instance Multi-Label Acoustic Classification of Plurality of Animals : birds, insects & amphibian

, , , , and .
Proc. 1st workshop Neural Information Processing Scaled for Bioacoustics -from neurons to Big Data-NIPS4B, page 164-174. NIPS4B, (December 2013)

Abstract

Nowadays, consulting firms on environment propose to evaluate impacts of transports and/or power production infrastructures on biodiversity using bioacoustic and adapted algorithms of signal processing. We present here our best algorithm (whose AUC score is 0.85\%). This is our contribution to the “Neural Information Processing Scaled for Bioacoustics ” (NIPS4B) workshop technical challenge 1 of NIPS 2013. Our objective was to obtain a bird-sound operational classification machine-learning model that environmental engineers (mostly or nithologists) could use to realise automatic inventories of acoustically active animals.

Tags

Users

  • @olivierlouis

Comments and Reviews