BeeSAM2: detecting bees in cherry flowers using timelapse images and foundational models
Devlin, J.; Macfarlane, F.; Karley, A.; Manfredini, F.; Williams, D.
Show abstract
Bees perform important pollination services in fruit crops such as cherry. Growers will often bring in bees to supplement natural pollinators. Monitoring the performance of these pollinators is important to understand the impact of augmenting pollinators on fruit yield particularly in relation to June drop which is a major cause of yield instability in the cherry industry. Timelapse imaging plus automated image analysis methods is a valuable tool in studying the role of bee pollination in fruit set. Timelapse cameras allow for continuous monitoring of the flowers, but manual analysis of the generated footage is very time consuming. We have developed a novel method of detecting bees in time lapse images, called BeeSAM2. This exploits both the zero shot detector Grounding Dino and the foundational model Segment Anything 2. Promising results are achieved with the method able to detect the bumblebee Bombus terrestris in images with a recall of 0.959 and precision of 0.991. These results are accurate enough to deploy our method to quantify bee activity in cherry plantations, advancing the ability of researchers to monitor bee interactions with flowers with a significant time saving over manual analysis of timelapse footage.
Matching journals
The top 3 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Tracking individual honeybees among wildflower clusters with computer vision-facilitated pollinator monitoring 97%
- Precise automatic classification of 46 different pollen types with convolutional neural networks 95%
- Utilizing CNNs for classification and uncertainty quantification for 15 families of European fly pollinators 94%
Similar papers in this journal
Similar papers in this journal
- An Assistive Computer Vision Tool to Automatically Detect Changes in Fish Behavior In Response to Ambient Odor 93%
- WormSwin: Instance segmentation of C. elegans using vision transformer 91%
- Computer vision and deep learning automates nocturnal rainforest ant tracking to provide insight into behavior and disease risk 91%
Similar papers in this journal
- An automated light trap to monitor moths (Lepidoptera) using computer vision-based tracking and deep learning 96%
- Ethoflow: computer vision and artificial intelligence-based software for automatic behavior analysis 93%
- High-Throughput and Accurate 3D Scanning of Cattle Using Time-of-Flight Sensors and Deep Learning 92%
Similar papers in this journal
- Towards edge processing of images from insect camera traps 95%
- Using photographs and deep neural networks to understand flowering phenology and diversity in mountain meadows 93%
- Using very high-resolution satellite imagery and deep learning to detect and count African elephants in heterogeneous landscapes 93%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.