核心概念
Automating Aedes aegypti egg counting using neural networks for disease control and research.
摘要
Aedes aegypti is a major disease vector, requiring accurate egg counting for control strategies.
Laborious manual egg counting can be automated using computer vision techniques.
Proposed dataset includes field and laboratory eggs, tested with three neural networks.
Faster R-CNN, SABL, and FoveaBox are applied to the task of egg counting.
Evaluation metrics include mAP, MAE, RMSE, precision, recall, and Pearson's coefficient.
FoveaBox outperformed other architectures in counting closely clustered eggs.
Challenges include high quantities of eggs, clusters, dirt presence, and perspective difficulties.
統計資料
ブラジルでは、デング熱のコストは2009年に5億1679万ドルから2013年、2015年、2016年、2019年には16億8830万ドルまで上昇した。
画像セットには123枚のフィールド卵と124枚のF1卵が含まれており、合計12.513個のA. aegypti卵がアノテーションされている。
引述
"The importance of counting eggs obtained in the field notwithstanding, there are research lines that require counting eggs laid in the laboratory."
"FoveaBox achieved better performance in counting extensive arrays of closely clustered eggs."