Example image where a bumble bee is automatically detected
by Devlin et al.
Bees are important for the pollination of flowers, and the subsequent fruit production, of many fruit crops such as sweet cherry. This means that fruit growers will often introduce commercially-produced bumblebees to their farms. Monitoring of these introduced bees is important to understand how bee activity affects fruit yield, particularly in relation to June drop which is a major cause of yield loss in cherry. Cameras can be used to continuously monitor flowers, but analysis of the generated footage is time consuming. The development of automatic image processing methods that can detect bees within camera images without manual image checking is a useful tool for studying bee pollination.
This work presents a novel method called BeeSAM2 for detecting bumblebees in time lapse images. This combines two previously developed methods Grounding DINO which detects objects in images when given a text prompt and Segment Anything 2 which tracks objects through videos when provided with an example object to follow. BEESAM2 detected over 95% of bumblebees captured in time-lapse images. This method is sufficiently accurate for counting bumblebees active in cherry flowers, advancing the ability of researchers to monitor bee behaviour and interaction with flowers, while saving significant time on video processing.
No comments:
Post a Comment