Abstract

This paper presents an approach to bee detection in video
streams using a neural network classifier. We describe the motivation
for our research and the methodology of data acquisition. The main
contribution to this work is a comparison of different color models used
as an input format for a feedforward convolutional architecture applied
to bee detection. The detection process has is based on a neural binary
classifier that classifies ROI windows in frames taken from video streams
to determine whether or not the window contains bees. Due to the type of
application, we tested two methods of partitioning data into training and
test subsets: video-based (some video for training, the rest for testing)
and individual based (some bees for training, the rest for testing). The
tournament-based algorithm was implemented to aggregate the results of
classification. The manually tagged datasets we used for our experiments
have been made publicly available. Based on our analysis of the results,
we drew conclusions that the best color models are RGB and 3-channeled
color models: RGB and HSV are significantly better than black & white
or the H channel from HSV.

Citations

Authors (2)