University of Twente Student Theses


Detecting stomata with computer vision for plant breeding

Vissers, W.K. (2023) Detecting stomata with computer vision for plant breeding.

[img] PDF
Abstract:Climate change is a growing concern and is affecting global food security. To address these issues, precision agriculture is emerging as a crucial research field. This field involves listening to the plants and meeting their direct needs. Stomata, the microscopic pores in plant leaves, play a vital role in regulating gas exchange, enabling the uptake of carbon dioxide and the release of oxygen during photosynthesis. The response of stomata to certain stimuli can be analysed using a microscope on live plants, with this knowledge better plants can be bread based on specific stomatal features. However, counting stomata is currently a time-consuming task that can be automated with the use of object detection models. The study focuses on automating this process using models from the yolov7 family, including small, medium, and large models. The study aimed to determine the best model based on various training methodologies and tested all models on unseen images from three different species. The results indicate that both the yolov7x (large) and yolov7-tiny (small) models performed well in detecting stomata. The yolov7x (large) model achieved a precision of 88.9% and recall of 75.1% on all test images, and a precision of 98.1% and recall of 95.5% on the most trained species, chrysant. On the other hand, the yolov7-tiny (small) model had a precision of 88.4% and recall of 80.3% on all test images, and a precision of 96.9% and recall of 95% on the chrysant test set. However, it is worth noting that the performance of the yolov7x model dropped when analysing tomato images, which contained the smallest stomata among all test images. Yolov7x proves to be a well capable model for the task of stomata detection, especially with chrysant. Future development should focus on increasing performance on tomato and cucumber.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:42 biology, 43 environmental science, 54 computer science
Programme:Creative Technology BSc (50447)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page