USING DEEP CNN ARCHITECTURES AND TENSORFLOW FOR BEE IDENTIFICATION ON IMAGES
DOI:
https://doi.org/10.34185/1991-7848.itmm.2023.01.095Keywords:
neural networks; tensorflow; deep learning; bees; computer vision; object detection; bees; CNN; single-shot detector; bee detection;Abstract
Bees are important pollinators for various plant species and are essential in maintaining our planet's biodiversity. Individual bee identification is an important task that, if performed well, will help to track the bee hive population and its health state in general. Photographs of bee hive frames with bees on them are being used as a data source for model training. These photographs were manually annotated using Remo annotation software. This study uses five widely used deep learning architectures (Faster R-CNN ResNet152, CenterNet ResNet50, SSD ResNet50, Faster R-CNN Inception ResNet, SSD MobileNet) for the purpose of bee detection on test images. These models were trained on the same dataset and evaluated on the same bee image set. According to the confusion matrix, SSD MobileNet architecture showed the best detection performance with 80% detected bees on test files, but it also had the falsest negative entries.
References
Garibalidi L. Wild pollinators enhance fruit set of crops regardless of honey bee abundance // Science. 2013., vip. 6127 T. 339. S. 1608-1611.
Wang J. The identification of butterfly families using content-based image retrieval // Biosystems Engineering. 2012., vip. 1 T. 111. S. 24-32.
Patel D., Bhatt N. Insect Identification Among Deep Learning’s Meta-architectures Using TensorFlow // International Journal of Engineering and Advanced Technology. 2019., vip. 1 T. 9. S. 1910-1914.
