Mitigating the taxonomic impediment problem of plants using ML and citizen science-Lebanon

PROJECT IMPACT

$3810

IEEE Funding

45

Estimated Impact

110

Students

5

Volunteers

Location: Lebanon
PROJECT LAUNCHED: March 2024
PROJECT LEADS:

Dr. Dany Abou Jaoude

Lebanon is a hot spot for vascular plant species, home to more than 2,600 vascular plant species. With such diverse flora, it can be difficult to properly identify species, and their respective needs. Because of this knowledge gap, these species have undergone a decline over the past few years. Plant databases in Lebanon are limited to specific nature reserves and only list the names of the species found in these reserves. To ensure better conservation, a bigger, more detailed database is needed. 

A team of students from American University Beirut took note of this issue, and with the Shouf Biosphere Reserve (SBR), are creating a real-time planet identification tool via a smartphone application. This tool is meant for use by nature reserve managers, conservationists, and nature enthusiasts, with the hope of creating a holistic database of species, areas, and seasons for Lebanese plants. The database will help stakeholders assess trends in species distribution over the years.

The team’s goal is to make their tool accessible, and to do so, plan to create a smartphone application that works with the camera already on the phone. This app would specifically be geared towards the flora of Lebanon, making it more effective for use in that biome than other apps already available. The app will even identify endangered species specifically, encouraging conservation. They will also draw attention to areas in Lebanon that need protecting to ensure that the natural flora can thrive. 

The team started by developing a preliminary plant dataset, using as much data as they could find online and in the literature.

The team then conducted pilot studies and workshops to educate and improve their project. Their workshops involved speakers, team members, students, volunteers, SBR staff, and field guides. Topics covered were taxonomy and database creation, primer to machine learning and computer vision, fieldwork and data collection, plant cultural significance, and biodiversity conservation.

Since May 2024, 110 student volunteers have conducted weekly field trips to collect plant images to populate the Lebanese plant species database. These images are being used for training the machine learning model for plant identification and classification purposes. So far, the training dataset contains 90 classes (plant species) with approximately 400 images per class. The model performs well, with around 95% accuracy on new and unseen plant images.

This project was made possible by $3,810.00 in funding from APS an EPICS in IEEE partner.

 

Contact Us

ADDRESS

Ashley Moran
445 Hoes Lane
Piscataway, NJ 08854, USA

PHONE

+1 732-562-6552

EMAIL

epicsinieee@ieee.org

CONNECT WITH US

JOIN OUR MAILING LIST