Bird Species Identification using Deep Learning

Abstract

These days some bird species are being found once in a while and whenever discovered arrangement of bird species forecast is troublesome. Normally, birds present in different situations show up in various sizes, shapes, shadings, and points from human viewpoint. Plus, the pictures present solid varieties to recognize the bird species more than sound grouping. Likewise, human capacity to perceive the birds through the pictures is more reasonable. So this strategy utilizes the Caltech-UCSD Birds 200 [CUB-200-2011] dataset for preparing just as testing reason. By utilizing deep convolutional neural organization (DCNN) calculation a picture changed over into grey scale arrangement to produce signature by utilizing tensor stream, where the numerous hubs of correlation are created. Test examination on dataset shows that calculation accomplishes an exactness of bird distinguishing somewhere in the range of 80% to more than 90%The trial study is finished with the Ubuntu 16.04 OS utilizing a Tensor stream library.