Abstract

Recent advancements in machine learning, in particular deep learning, enabled automatic object-classification based on visual information with high accuracy. These methods also make the automatic detection and monitoring the state (e.g. health) of animals much more feasible, which presents exciting opportunities for collaboration of biologists and veterinary researchers.

During this talk, we will present the hardware and software solutions we have designed and/or implemented for automatic disease detection. The sensory platform we built collects multimodal imaging data, including far infrared thermal and near infrared night vision images. The platform is based on off-the-shelf components and open-source software, which offers great adaptability for other monitoring purposes, including wildlife camera traps, on a cost effective basis. 

Our background is not veterinary science, but in sensory biology and psychophysics; therefore we will recount our transition from working on camouflage to the veterinary world and present examples from our work in veterinary image diagnostics, including automatic monitoring of cattle and horses.

 

Biography

John and Laszlo are joint fellows on an EPSRC Innovation Fellowship project that will conclude this summer, after which they both start lectureships at Bristol Vet School. John’s background is in psychology, including visual perception, uncertainty and behavioural decision making. Laszlo’s background is in sensory biology with a focus on animal and military camouflage. They both have completed their graduate and postgraduate degrees at University of Bristol.

Laszlo Talas

John Fennell