2020/07/21 by Mona Minakshi, Pratool Bharti, Minakshi, Mona +7
Computer Science · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #Digital Imaging for Blood Diseases #FOS: Computer and information sciences #Mosquito-borne diseases and control
paper · pdf · doi:10.48550/arxiv.2007.11052
openalex publication_date 2020/07/21 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
We design a framework based on Mask Region-based Convolutional Neural Network\n(Mask R-CNN) to automatically detect and separately extract anatomical\ncomponents of mosquitoes - thorax, wings, abdomen and legs from images. Our\ntraining dataset consisted of 1500 smartphone images of nine mosquito species\ntrapped in Florida. In the proposed technique, the first step is to detect\nanatomical components within a mosquito image. Then, we localize and classify\nthe extracted anatomical components, while simultaneously adding a branch in\nthe neural network architecture to segment pixels containing only the\nanatomical components. Evaluation results are favorable. To evaluate\ngenerality, we test our architecture trained only with mosquito images on\nbumblebee images. We again reveal favorable results, particularly in extracting\nwings. Our techniques in this paper have practical applications in public\nhealth, taxonomy and citizen-science efforts.\n