The aim is to build an image classification model allowing to predict species of insects from pictures. The dataset consists in 20348 pictures of insects from different species gathered from the SPIPOLL project and labeled. There are 18 different classes each one corresponding to a different insect specie. Each picture is a 64x64 colored image, which makes a total of 64×64×3=12288 features per picture.
Reference : http://www.spipoll.org/