Classification of Mice Head Orientation Using Support Vector Machine and Histogram of Oriented Gradients Features

The development of computational tools allows the implementation of new technologies, promoting an advance in Neuroscience research and raising the limits of experimental design. The tool proposed in this study is initially described with the acquisition of the images, then performing the animal tra...

Full description

Saved in:

Bibliographic Details
Published in:	2018 International Joint Conference on Neural Networks (IJCNN) pp. 1 - 6
Main Authors:	Menezes, Richardson Santiago Teles de, Lima, Lucas de Azevedo, Santana, Orivaldo, Henriques-Alves, Aron Miranda, Cruz, Rossana Moreno Santa, Maia, Helton
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01-07-2018
Subjects:	Feature extraction Head Image edge detection Magnetic heads Mice Support vector machines
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	The development of computational tools allows the implementation of new technologies, promoting an advance in Neuroscience research and raising the limits of experimental design. The tool proposed in this study is initially described with the acquisition of the images, then performing the animal tracking, extraction of characteristics, and finally the classification stage. Support Vector Machines are known as one of the most powerful supervised learning algorithms widely used for multiclass image classification. In this work, we propose the classification of mice head orientation using a support vector machine trained with Histogram of Oriented Gradients descriptors, extracted from behavioral video records. We analyze a dataset composed of 1800 images, using 80% of the images for training and 20% for tests. The classifier admits 8 outputs, represented by the trigonometric circle divided into 45 degree parts, allowing the classification of mice head orientation in 8 possible directions. The results show that the mean accuracy reached by the classifier is 99.44%, with confounding errors only in neighboring classes. The computational development was performed using the C++ programming language and the Open Source Computer Vision Library. In this way, considering the accuracy achieved by the classifier and its well-known low computational cost after training, it is possible with this tool to build from simple behavioral experiments of social interaction to the synchronization of the real-time response of the animal's head orientation with light stimuli needed in real-time optogenetic experiments.
ISSN:	2161-4407
DOI:	10.1109/IJCNN.2018.8489558