Periodic Reporting for period 1 - AI4AGRI (Romanian Excellence Center on Artificial Intelligence in Earth Observation Data for Agriculture)
Période du rapport: 2022-10-01 au 2025-09-30
The AI4AGRI project aimed at creating a dedicated research center for AI in EO for the agricultural sector. AI4AGRI research center has become a reference to train young scientists in the domain of AI for agriculture, providing maps of vegetation status for Romanian farmers. AI4AGRI developed administrative and management skills for research and innovation. Improved excellence capacity and resources was achieved through 9 partner visits and 10 staff exchanges, exceeding the planned minimum of 4 for both. Development of young researchers was realized with 12 short-term training sessions, 2 summer schools and 7 workshops, surpassing initial plans. Enhanced reputation and research profile were reached with 16 scientific papers in open-access journals and 28 in conferences, exceeding targets. Strengthened research capacity was developed during the project with 4 joint PhD programs and 6 collaboration agreements, surpassing planned numbers. Creativity in AI and EO applications was fostered through collaborative efforts across all work packages.
WP1 focused on enhancing the research capacity of UTBV through learning best practices and elaborating a scientific roadmap. 9 partner visits and 9 staff exchanges were conducted, facilitating knowledge transfer from project partners to UTBV. The scientific roadmap identified key research areas in AI and Agriculture. A research capacity evaluation was conducted towards the end of the project, showing significant improvements in the team's capabilities.
In WP2 a total of 12 short-term training sessions were completed. Senior team members delivered presentations to ESRs, with weekly technical meetings and ad-hoc discussions, enhancing their knowledge. 2 summer schools were successfully organized. In total, there were 86 participants from more than 10 countries participating to the events, in addition to the ESRs of the UTBV team. 7 workshops were organized, that attracted tens of participants each. The training materials and 2 support materials for the summer schools were published as open-access material on Zenodo. A textbook on AI (in Romanian) for ESRs and students was published.
WP3 established key research directions, including spectral image visualization, soil roughness estimation, crop early identification etc. 5 datasets were created for testing and validation, made available on Zenodo as open access. A partnership with the Field Data Zoom company from Bucharest, Romania was formed to integrate developed applications into commercial tools. A a free ESA SNAP plugin for the visualization of spectral images was developed and provided in open access on Zenodo. The main results are new approaches and algorithms, with Matlab or Python implementation, for performing various tasks in the EO data analysis chain.
For WP4 six collaboration agreements were signed with various stakeholders, enhancing project outreach. Multiple project proposals were submitted, with several accepted for funding, ensuring continued research and development efforts. Various top institutions were approached.
In WP5 a plan for dissemination was carefully designed to address the best publication venues in remote sensing and connected domains. 16 journal papers published in high-impact journals and 28 conference publications, all open access, with significant participation in international events. The project implemented various communication strategies: logo, website, the official opening of the AI4AGRI Excellence Center in June 2023, 15 press releases and 7 articles published in newsletters and magazines to promote project activities. Participation in outreach events like World Space Week, engaging over 600 pupils and professors, and the European Researchers' Night.
The AI4AGRI project has made strides in delivering scientific and societal impacts through research and community engagement: 2 PhD theses completed, with additional career advancements for 3 team members, 10 master dissertations and internships conducted, farmers received NDVI maps for free, enhancing their ability to interpret satellite data.
In WP2 a total of 12 short-term training sessions were completed. Senior team members delivered presentations to ESRs, with weekly technical meetings and ad-hoc discussions, enhancing their knowledge. 2 summer schools were successfully organized. In total, there were 86 participants from more than 10 countries participating to the events, in addition to the ESRs of the UTBV team. 7 workshops were organized, that attracted tens of participants each. The training materials and 2 support materials for the summer schools were published as open-access material on Zenodo. A textbook on AI (in Romanian) for ESRs and students was published.
WP3 established key research directions, including spectral image visualization, soil roughness estimation, crop early identification etc. 5 datasets were created for testing and validation, made available on Zenodo as open access. A partnership with the Field Data Zoom company from Bucharest, Romania was formed to integrate developed applications into commercial tools. A a free ESA SNAP plugin for the visualization of spectral images was developed and provided in open access on Zenodo. The main results are new approaches and algorithms, with Matlab or Python implementation, for performing various tasks in the EO data analysis chain.
For WP4 six collaboration agreements were signed with various stakeholders, enhancing project outreach. Multiple project proposals were submitted, with several accepted for funding, ensuring continued research and development efforts. Various top institutions were approached.
In WP5 a plan for dissemination was carefully designed to address the best publication venues in remote sensing and connected domains. 16 journal papers published in high-impact journals and 28 conference publications, all open access, with significant participation in international events. The project implemented various communication strategies: logo, website, the official opening of the AI4AGRI Excellence Center in June 2023, 15 press releases and 7 articles published in newsletters and magazines to promote project activities. Participation in outreach events like World Space Week, engaging over 600 pupils and professors, and the European Researchers' Night.
The AI4AGRI project has made strides in delivering scientific and societal impacts through research and community engagement: 2 PhD theses completed, with additional career advancements for 3 team members, 10 master dissertations and internships conducted, farmers received NDVI maps for free, enhancing their ability to interpret satellite data.
The scientific roadmap identified new research directions and emerging topics like: AI-based visualization of spectral images, NDVI map coloring, crop rotational model-based identification, use of electromagnetic models for AI training. Innovative AI-based approaches included the soil roughness estimation using digital images, data fusion techniques for combining radar and optical data to improve crop identification, early crop identification and exploration of vegetation indices as data descriptors for crop identification and segmentation. New datasets published as Open Science contributions: Soil Roughness, DACIA5 for crop identification using Sentinel-1 and 2 data and HyDACI6 – a unique PRISMA-based hyperspectral data set for crop identification. DACIA5 was integrated into international challenges like ImageCLEF. Integration into operational tools of the AI-based visualization of spectral data through the development of a plugin for ESA SNAP. Publication of 16 open-access journal papers and 26 conference papers in high-impact venues, advancing knowledge in AI and EO applications for agriculture. Provision of NDVI maps to farmers for free, increasing awareness on the EO and AI-based tools and enabling better interpretation of satellite data for crop monitoring and management.