Computer Vision Phenomics for Insect Traits: Predicting Age and Weight of Pollinators Using Convolutional Neural Networks
(2025) BINP52 20242Degree Projects in Bioinformatics
- Abstract
- Pollinators such as Bombus terrestris play a vital role in ecosystem stability and agricultural productivity. Developing methods to estimate key biological traits such as age and weight is crucial for advancing ecological understanding and conservation efforts. In eusocial insects like B. terrestris, age and weight influence behaviour, task allocation, and survivability, but traditional methods for estimating these traits are often lethal, labour intensive, or lack precision. This study investigates whether morphological features of the wing, including venation patterns and wing wear, can be used to predict individual traits. As part of this study multiple convolutional neural networks (CNNs) were trained on bee wing images with known age... (More)
- Pollinators such as Bombus terrestris play a vital role in ecosystem stability and agricultural productivity. Developing methods to estimate key biological traits such as age and weight is crucial for advancing ecological understanding and conservation efforts. In eusocial insects like B. terrestris, age and weight influence behaviour, task allocation, and survivability, but traditional methods for estimating these traits are often lethal, labour intensive, or lack precision. This study investigates whether morphological features of the wing, including venation patterns and wing wear, can be used to predict individual traits. As part of this study multiple convolutional neural networks (CNNs) were trained on bee wing images with known age or weight. The models learned to identify patterns associated with these traits, offering a new method for high-throughput, non-invasive trait estimation using wing images alone. The evaluation of the developed models on unseen wing images suggests that wing images can be used as a practical alternative to traditional estimation methods that outperform statistical methods. (Less)
- Popular Abstract
- Computer Vision Phenomics for Phenological Analysis
Bees play an essential role in our lives. They pollinate many of the crops we eat and help keep ecosystems in balance. But their populations are under pressure, facing threats from disease, pesticides, and climate change. To protect them, we need to better understand how bees respond to their environment. One important factor is the agestructure of bees in a colony. Age matters because young and old bees play different roles, from cleaning and nursing to foraging for food. Until now, identifying the age of a bee has required time-consuming observations or invasive methods.
My project takes a different approach by asking the question if wings of a bee hold enough clues to infer its... (More) - Computer Vision Phenomics for Phenological Analysis
Bees play an essential role in our lives. They pollinate many of the crops we eat and help keep ecosystems in balance. But their populations are under pressure, facing threats from disease, pesticides, and climate change. To protect them, we need to better understand how bees respond to their environment. One important factor is the agestructure of bees in a colony. Age matters because young and old bees play different roles, from cleaning and nursing to foraging for food. Until now, identifying the age of a bee has required time-consuming observations or invasive methods.
My project takes a different approach by asking the question if wings of a bee hold enough clues to infer its age. Just like wrinkles tell us something about humans, the tiny details in bee wings give indications of their age. To test this, I combined wing images from different sources into a single dataset. I then trained convolutional neural networks (CNNs) to search for patterns assosiated with age.
The resulting models were able to estimate the age of bees just by an image of their wings. This opens up exciting possibilities. Beekeepers and researchers could use wing images to quickly and non-invasively assess the age of bees and the age structure of a colony. This information would help them detect early signs of stress or imbalance, such as too few young bees to take care of larvae. Over time, such tools could support healthier colonies, which in turn benefits agriculture and biodiversity.
Master’s Degree Project in Bioinformatics 60 credits 2025
Department of Biology, Lund University
Advisor: Charlie Nicholson
Department of Biology (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9212649
- author
- Materna, Jakob
- supervisor
-
- Charlie Nicholson LU
- Ola Olsson LU
- organization
- course
- BINP52 20242
- year
- 2025
- type
- H2 - Master's Degree (Two Years)
- subject
- language
- English
- id
- 9212649
- date added to LUP
- 2025-09-22 13:55:37
- date last changed
- 2025-09-22 13:55:37
@misc{9212649,
abstract = {{Pollinators such as Bombus terrestris play a vital role in ecosystem stability and agricultural productivity. Developing methods to estimate key biological traits such as age and weight is crucial for advancing ecological understanding and conservation efforts. In eusocial insects like B. terrestris, age and weight influence behaviour, task allocation, and survivability, but traditional methods for estimating these traits are often lethal, labour intensive, or lack precision. This study investigates whether morphological features of the wing, including venation patterns and wing wear, can be used to predict individual traits. As part of this study multiple convolutional neural networks (CNNs) were trained on bee wing images with known age or weight. The models learned to identify patterns associated with these traits, offering a new method for high-throughput, non-invasive trait estimation using wing images alone. The evaluation of the developed models on unseen wing images suggests that wing images can be used as a practical alternative to traditional estimation methods that outperform statistical methods.}},
author = {{Materna, Jakob}},
language = {{eng}},
note = {{Student Paper}},
title = {{Computer Vision Phenomics for Insect Traits: Predicting Age and Weight of Pollinators Using Convolutional Neural Networks}},
year = {{2025}},
}