The reduction of natural habitats has caused rising tensions between Humans and elephants in Africa. One side effect of this is the raiding of farmers crops by elephants which when combined with poverty can lead to unfortunate consequences. Our project aims to use low-cost hardware and smart imaging to support species co-existence.
The first step was to source some images of elephants to train the deep learning network. A collection of images was found here: https://www.kaggle.com/alessiocorrado99/animals10/data
However, when examining the elephant data - many images had watermarks, some actually included other images such as rhinos and drawings of elephants. The dataset was then processed to include only __ of these images.
The images were then labelled using Matlab's inbuilt image labeler application.
To train the network, a high powered Graphics card was purchased. The code used to train the network was written in Matlab 2019b alongside the CUDA toolkit 10.0.
Comments