Latest Bloggers

Doris | SPOTTERON
10 posts
Design and communication at SPOTTERON
Sylvain Poggi (INRAE)
3 posts
Quantitative ecologist and data scientist at th...
Sara Boschi (Unisi)
1 post
Sara Boschi (Unisi) has not set their biography yet
Agostino Strangi (CREA - DC)
4 posts
Agostino Strangi (CREA - DC) has not set their biography yet

From Twitter

05 August 2024
Pest management
#IPMPopillia
Pupils from class 2D had a good intuition! In fact, the Fungzuka device is being tested by CREA's researchers since summer 2023. The "Fungzuka" is an "Attract-Infest-Release" device that is conceived to bring adult Popillia japonica insects in c...
10 July 2024
Citizen Science
#IPMPopillia
Monitoring
In recent months, the IPM Popillia Consortium has collaborated with SPOTTERON to create informational and dissemination materials about the invasive Japanese beetle (Popillia japonica), which has been spreading across Europe for several years. R...
10 June 2024
Citizen Science
#IPMPopillia
Prof. Francesco Nardi (University of Siena) and Prof. Rossella Annoni (junior high school G. Falcone, Cassina de' Pecchi, Milan) have been working together, this past year, with class 2D (12-13 years old pupils) on Popillia within the context of the ...
IPM_Popillia-image-for-page

Wrapping up the Project: Achievements have been made by PESSL

Pessl Instrument GmbH(PESSL)

 Pessl Instruments GmbH has been producing reliable measuring instruments for more than 40 years, developing a range of devices for measuring and monitoring various agricultural parameters. As one of the leading IoT providers in agriculture, PESSL offers innovative, cost-effective solutions for more efficient farm management. Participation in the IPM-Popillia project provides PESSL with the opportunity to bring its expertise in sustainable agriculture to farmers, offering high-value, customized, cost-effective solutions while contributing to global environmental protection. 

 The IPM-Popillia project, funded by @EU Horizon 2020 research and innovation program, started in September 2020. Our expert team has developed a monitoring device aimed at detecting and monitoring P. japonica. The insect traps include electronic devices with 10 MP lenses on the top of the housing and are self-sufficient, powered by a battery and solar panel. The traps are equipped with sensors collecting climatic data, such as temperature and relative humidity which will be used to get more detailed insight into the flight behavior of P. japonica. In addition, the trap system is equipped with a lure, attracting the target species to enter the trap system. After entering the trap, insects get fixed and photographed. The photos serve as a base for the development of an automatic detection tool specifically for P. japonica. Using deep learning systems and artificial neural networks, the system is trained to accurately detect and differentiate the target insect from non-target species

ACHIEVEMENTS IN A NUTSHELL

  • Prototypes were collected from the field, upgraded with climatic sensors measuring relative humidity and temperature, and updated with the latest firmware.
  • These prototypes have been installed in 6 countries (Portugal, Italy, Switzerland, France, Austria, and Slovenia) across 14 locations.
  • The installations took place in low to moderately infested areas, to capture as many photos as possible to further train the insect identifier. However, the system was designed to be used for surveilling the first occurrence in newly invaded areas.
  • The device's firmware, control unit, and insect recognition capabilities have been optimized for stability and accuracy.
  • The system is adaptable to various network conditions. If strong connectivity is unavailable, it can automatically switch to 2G or 3G networks, ensuring continued data transmission despite reduced image quality.
  • The alert message function is implemented on the iSCOUT Gallery page for P. Japonica.
  • A new identification model for species has been released to production in the summer of 2024, with a second version following in autumn 2024.

THE LATEST P. Japonica IDENTIFICATION MODEL IN PRODUCTION

Significant advancements have been made in detection model development in the past 26 months: the genus-level identification model was released in July 2023. An initial release of the species-level identification model was made in May 2024, followed by a subsequent upgrade in November 2024, with a refined species-level identification model that reached 98.3% accuracy.

The model's performance has been enhanced through the labeling, evaluation, and application of photos from the collected data. Also, the use of dead specimens continued by placing them on glue boards and photographing them under various conditions (e.g., lighting, focus, positioning) to create a more diverse dataset.

Model Performance: Insects Detection

The object detection phase involves identifying insects on sticky plate images and classifying them by order. The performance of our detection models is summarized below:

Taxonomic rank
Model Version
Release Date
Overall mAP*
Order A3
10 29.05.2024
0.454
*mean Average Precision

Model Performance: P. Japonica Identification

For the P. Japonica insect species, additional classification models were trained and evaluated. These models achieved excellent results, as outlined in the table below:

Taxonomic Rank
Release Date
P. Japonica F1 Score*
Overall Accuracy**
Species B2
19.11.2024 0.9956 0.9830

*F1 score

**Overall Accuracy of all the insect labels included in the whole model

Training datasets for the P. Japonica species were developed iteratively through continuous trials and image collection, resulting in models capable of reliably identifying this target species.

Last but not least, We are delighted to announce the successful completion of all project tasks, deliverables, and milestones as defined in the DOA for Pessl Instruments.

We thank again all the project partners for their dedicated support throughout the field activities.

ABOUT THE AUTHORS:

M.Sc. Junia Rojic, Project support officer

At Pessl Instruments, I am responsible for the administrative tasks related to projects, including among others proposals, reports, deliverables, time schedules, meetings, minutes, and budget control. I studied Industrial Engineering at the Federal University of São Carlos, Brazil, recognized in Austria as a Master's degree in Materials Science. With more than 15 years of experience in Project Management, including some years in Brazil working at R&D and IT companies, I am delighted to be part of a European Project, with many important stakeholders, and most importantly, supporting agriculture and food production in the 21st century. In my free time, I do sports, such as volleyball, but also enjoy spending some time reading and in nature hiking.

Damir Najvirt, Machine Learning Engineer

I studied physics, specializing in stochastic processes and their role in the development of financial markets at the Faculty of Natural Sciences and Mathematics in Maribor, Slovenia. I have been working for Pessl Instruments since 2016, where I developed image processing solutions for our camera products and automated systems for fruit and insect recognition. I am responsible for designing, evaluating, and enhancing our machine learning stack. I am especially passionate about the transformation of data into valuable information that, in turn, guides farmers all over the world.

Minkyeong Kim, Project Assistant

As a project assistant at Pessl Instruments, I provide administrative support for both internally and externally funded projects, including dissemination, communication activities and preparation of documents and reports. I graduated from Sookmyung Women's University in Seoul, Korea with a degree in Business Administration and Economics. With diverse experiences in Finance and HR departments, my communication skills are adaptable and borderless. I am passionate about the agricultural field and its potential and I'm thrilled to be part of the team PESSL. In my free time, I love to explore new cultures through travel and photography.

General Assembly of the IPM Popillia Project in Po...

By accepting you will be accessing a service provided by a third-party external to https://www.popillia.eu/

App Download Links (QR-Codes)

Android
(Google Play) 

qrcode app Popillia googleplay

iPhone
(Apple App Store) 

 qrcode appPopillia appstore

 EU Flag This project has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement No 861852

twitter