Counting cattle in UAV images using convolutional neural network

Video dataset of sheep activity for animal behavioral analysis via deep learning

“…A comprehensive video set of sheep activities were captured via ground-based cameras and prepared in this article, which can be used for developing new deep learning models and enabling accelerated application of pretrained models for other similar domains. However, we note that, since herd management via Unmanned Aerial Vehicles (UAV) (including drone technologies [ 1 , 8 ] and networks using drone captures to detect and count animals [9] ) have recently gained interest, the accuracy of activity detection techniques using this proposed dataset may vary compared to the datasets obtained from UAV videos, due to the differences between the learned features of the deep learning models. The above technical barrier may, however, be overcome by using Transfer Learning approaches (transferring model parameters and utilizing minimal or small datasets) to obtain higher-accuracy detections.…”

Use of drone technology in animal health and management

“…UAVs, in combination with the proposed approach, improve herd assessments and farm management by automating cattle counting. The study by De Lima Weber et al [32], using Convolutional Neural Networks and YOLOv4, YOLOv5 models resulted in high accuracy in cattle counting and suggested the use of larger data sets by considering various factors.…”

Video Dataset of Sheep Activity for Animal Behavioral Analysis Via Deep Learning

“…A comprehensive video set of sheep activities were captured via ground-based cameras and prepared in this article, which can be used for developing new deep learning models and enabling accelerated application of pretrained models for other similar domains. However, we note that, since herd management via Unmanned Aerial Vehicles (UAV) (including drone technologies [ 1 , 8 ] and networks using drone captures to detect and count animals [9] ) have recently gained interest, the accuracy of activity detection techniques using this proposed dataset may vary compared to the datasets obtained from UAV videos, due to the differences between the learned features of the deep learning models. The above technical barrier may, however, be overcome by using Transfer Learning approaches (transferring model parameters and utilizing minimal or small datasets) to obtain higher-accuracy detections.…”