A CNN-based identification of honeybees' infection using augmentation

Honeybees are one of the most important pollinators and crucial for our survival on earth. In recent years, bee colonies have been under continuous threat of the parasitic mite varroa destructor. Designing an efficient system to detect varroa infestation has been a formidable challenge and a topic of intense inter-est among academics. To develop a robust deep neural network classification model, high-quality and comprehensive image data is required. Collecting and preparing such data is laborious, time-consuming and expensive in a beehive environment. This study presents a computer vision-based method for classifying infected and healthy honeybees, considering imprecise and insufficient visual data. As a remedy, picture contrast enhancement and GAN-based image data augmentation techniques are applied to optimise data quantity and quality, consecutively. The binary classification is performed using an optimised CNN method. The image enhancement technique not only improved the sharpness of the image but also boosted the CNN classifier performance. The optimised CNN performance improved significantly with GAN-based augmented data compared to both original and conventionally augmented data. Experimental results of the proposed model demonstrate encouraging results for identifying varroa mites on images which paves the way for future beehive monitoring systems.