Bee keeping has gained quite an attention by local Malaysian especially towards a particular bee species known as Meliponine, which is a tribe of stingless bee. Up until now, the identification of this species still heavily relies on the input of a handful key experts. Many studies have shown that the honey contains beneficial nutrients that can combat against various diseases, which makes it very popular and valuable as compared to the normal honey. This study aims to aid the process of bee keeping by developing a visual-based guidance system.The main process of this system will involve image segmentation. Due to the miniature size of the Meliponine, the conventional way of image segmentation using background subtraction is becoming irrelevant and difficult. In recent years, the convolutional neural network (CNN) has been gaining huge ground in object detection due to its high accuracy and efficiency. In this study, we fine-tuned the Faster R-CNN architecture to perform segmentation on the Meliponine image from its background. Our dataset consists of 400 image frames from videos collected in local Meliponine farm in Malaysia. On these datasets we achieved 74% accuracy, which seems promising for further study. We also incorporated the Mask R-CNN architecture to carry out instance segmentation as opposed to the bounding box level detection provided by Faster R-CNN. Using Mask R-CNN, the computing time for training significantly improved.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.