School of Computing

Dec 11
14:00 - 15:00
compint: Alex (Aram) Ter-Sarkisov
Computational Intelligence Group Seminar
Dense pixel-level segmentation of instances of cows in a farm for early lameness detection

Abstract: 

We present a novel deep learning algorithm for cow segmentation that is a first step towards cow behavior classification and early identification of lameness. We train a fully convolutional neural network (FCNN) that learns to segment a cow throughout a video. A database for training is constructed from a raw video using naive segmentation on which FCNN is trained. Our approach is particularly useful for environments with heavy occlusion and high degree of similarity between segmented objects.

Location

SW101,
Cornwallis South West,
University of Kent,
Canterbury,
Kent,
CT2 7NF
United Kingdom
Map

Details

Contact: Marek Grzes
E: m.grzes@kent.ac.uk

School of Computing, University of Kent, Canterbury, Kent, CT2 7NF

Enquiries: +44 (0)1227 824180 or contact us.

Last Updated: 14/08/2015