AI4AGRI website report
As the AI4AGRI Centre of Excellence concludes its initial funded phase, we reflect on three transformative years that elevated Romania’s research capabilities in AI and Earth Observation (EO) for agriculture.
Project Goals and Vision
Launched in 2022 and hosted at Research and Development Institute of Transilvania University of Brașov, the AI4AGRI Romanian Excellence Center on AI for Agriculture goal is to:
- Align UTBV’s research with EU excellence.
- Promote the next generation of researchers.
- Enhance the national and international visibility of UTBV.
- Develop administrative and technical capacity for high-impact research.
- Apply AI and EO to real-world agricultural challenges.
Thanks to partnerships with institutions in France and Italy and active engagement with farmers and stakeholders, we are proud to share our most impactful outcomes.
Developing a network of EO expertise for AI assisted agriculture
AI4AGRI brought together a dynamic network of researchers, collaborating across the three involved countries:
Romania
- Andreea NITU, PhD student, female
- Artur KAZAK, PhD student, male
- Andrei RACOVITEANU, PhD, male
- Matei DEBU, MSc student, male
- Adrian RUJOI, junior researcher, male
- …
Italy
- Giuliano Lorenzo Papale, PhD Student, male
- Giorgia Salvucci, Post-Doc Researcher, female
- Ilaria Petracca, Post-Doc Researcher, female
- Giorgia Guerrisi, Post-Doc Researcher, female
- …
France
- Josiane Mothe, Professor, female
- Olivier Teste, Professor, male
- Max Chevalier, Professor, male
- Moncef Garouani, Assistant professor, male
- Lotfi Chaari, Professor, male
- Pape Ibrahima THIAM, PhD student, male
- Serge Molina, Research Engineer, male
- Mohammad El Sakka, PhD Student, male
Examples of Collaborations and Impacts
New National and International Networks
The project strengthened UTBV’s collaborations with over 25 institutions across Europe such as Tor Vergata University (Italy), the Joint Research Centre (JRC) at Ispra (Italy), and CERN-NTN doctoral program (Norway).
UTBV also joined EARSeL, Copernicus Academy, and expanded cooperation with national agencies like the Romanian Space Agency and ESRI Romania.
Projects and Technology Transfer
Since 2024, the center has acquired over €430k in research funding, coordinating or participating in:
- ELIAC: Early crop identification for Romanian farmers (with OGOR)
- AICoRS: AI for radiation sensing on satellites
- IMINT: AI-based EO for detecting small ground objects
- SEEN: AI and social engagement in entrepreneurship
Dissemination and Leadership
Our researchers published in leading journals including IEEE Transactions on Geoscience and Remote Sensing and European Journal of Remote Sensing. The Center organized and chaired sessions at major international events such as IGARSS, WHISPERS, or ISSCS, and contributed to special issues and scientific book chapters.
Challenges Addressed
At the start of AI4AGRI, the Steering Committee identified areas needing attention:
- Gender balance in the team
- Lack of joint PhD supervision and IP activity
- Few industrial partnerships
Today, we are proud to report measurable progress:
- Increased female participation in the team
- Two joint PhD programs (with Italy and Moldova) and one more in progress
- Three ongoing projects with industry involvement
- Two patents in preparation and new career advancements among team members
Examples of Scientific Advances and Highlights
Publications (non exhaustive list)
More than 30 research works have been published through the AI4AGRI funding phase, covering the following topics related to AI, EO and their application to agriculture:
Spectral & Multi‑Source Data Visualization
- Spectral Image Data Fusion for Multisource Data Augmentation
- Multisource Remote Sensing Data Visualization using Machine Learning
- AI-Based Visualization of Remotely-Sensed Spectral Images
- Hyperspectral Image Visualization Based on Maximum-Reflectance Wavelength Colorization
Soil Roughness & Fractal Analysis
- Deep automatic soil roughness estimation from digital images
- Soil Roughness Estimation Using Fractal Analysis on Digital Images of Soil Surface
- Multispectral Fractal Image Analysis for Soil Roughness Estimation at Various Altitudes
- Convolutional neural network hardware implementation for soil roughness estimation
Crop Monitoring & Vegetation Indices
- NDVI Computation from Hyperspectral Images
- A Review of CNN Applications in Smart Agriculture Using Multimodal Data
- Artificial Intelligence to Advance Earth Observation: A review
- Images and CNN applications in smart agriculture
Edge‑AI Applications & Forest Monitoring
- Smart Weed Control: Real‑Time Inference and On‑board Data Processing using Edge‑AI
- Annotating Satellite Images of Forests with Keywords from a Specialized Corpus in the Context of Change Detection
Datasets Released
- DACIA5: Sentinel‑2 multispectral dataset for crop identification over Brașov area, Romania
- muDACIA5: Sentinel‑2 time series dataset (2020–2024) for crop monitoring in Brașov
- HyDACI6 PRISMA dataset: Hyperspectral PRISMA-based data over Brașov for crop identification
- Soil roughness estimation dataset: Digital image collection labeled for soil roughness metrics
- Multi‑spectral fractal images dataset: Fractal-based labeled spectral data for analysis tasks
- AgriPotential – Multi‑temporal multispectral dataset for agricultural mapping in Southern France
AI4AGRI website report

AI4AGRI website visitors location distribution
Since its launch, the AI4AGRI website has attracted significant attention from a diverse, international audience interested in agricultural innovation and AI research.
Website stats
- 3,35K visitors have explored the website, demonstrating growing interest and engagement within the community.
- The website recorded 7,62K page views, indicating that visitors are actively browsing multiple pages per session.
International reach
The website has drawn visitors from 78 countries, highlighting AI4AGRI’s global appeal.
The visitor count for AI4AGRI members countries are:
- Romania: 587 visitors
- France: 205 visitors
- Italy: 130 visitors
Newsletters
To keep the community informed and connected, AI4AGRI has published 15 newsletters, each:
- Highlighting a researcher which takes part in the project
- Presenting news in which the project is involved
- Showcasing publications made as part of the AI4AGRI project
- Sharing related news and events
- Featuring a selected paper relevant the the AI and EO assisted agriculture community
Latest AI4AGRI Publications

Adaptive Select Loss Strategy for Semantic Segmentation of Agricultural Crop Images
🧑 Corneliu Florea, Laura Florea, Mihai Ivanovici
📅 July 2025
We address the problem of agricultural image segmentation by introducing a novel loss formulation called Adaptive Select Loss (ASL), inspired by the Top-k loss strategy. While Top-k loss was originally designed for classification tasks, ASL is specifically tailored for semantic segmentation. It exploits the hierarchical structure of loss computation specific in semantic segmentation – aggregated first at the pixel level, then at the image level – while accounting for the imbalance between precise but scarce image-level annotations and noisy yet abundant pixel-level labels. ASL selectively aggregates loss from a “Few” (Top-k) of the most informative image-level instances and from “Almost all” (remove few) pixel-level data, thereby balancing robustness and sensitivity to noise. To ensure stability during training, we introduce a derivative smoothing mechanism that addresses the convergence issues introduced by the hard selection threshold, particularly when training with small number of images for loss aggregation. Empirically, the proposed approach improves boundary localization and segmentation quality in the presence of annotation noise. We evaluate ASL on three challenging semantic segmentation tasks—two agricultural and one mixed—using a visual transformer backbone, including hyperspectral data. ASL achieves consistent performance improvements, with gains of approximately 2.5% on hyperspectral and satellite imagery, and up to 6% on RGB-D data plant segmentation problem.

NDVI and Beyond: Vegetation Indices as Features for Crop Recognition and Segmentation in Hyperspectral Data
🧑 Andreea Nițu, Corneliu Florea, Mihai Ivanovici, Andrei Racoviteanu
📅 June 2025
Vegetation indices have long been central to vegetation monitoring through remote sensing. The most popular one is the Normalized Difference Vegetation Index (NDVI), yet many vegetation indices (VIs) exist. In this paper, we investigate their distinctiveness and discriminative power in the context of applications for agriculture based on hyperspectral data. More precisely, this paper merges two complementary perspectives: an unsupervised analysis with PRISMA satellite imagery to explore whether these indices are truly distinct in practice and a supervised classification over UAV hyperspectral data. We assess their discriminative power, statistical correlations, and perceptual similarities. Our findings suggest that while many VIs have a certain correlation with the NDVI, meaningful differences emerge depending on landscape and application context, thus supporting their effectiveness as discriminative features usable in remote crop segmentation and recognition applications.

Comparing Blind Image Quality Metrics for Reliable Image Assessment
🧑 Cristian George Fieraru, Maria Biserică, Ioana Cristina Plajer, Mihai Ivanovici
📅 June 2025
Reliable image quality assessment is essential not only in digital photography but also as a key metric for evaluating the performance of algorithms and models designed for image quality enhancement or generation. In recent years, a wide range of image quality assessment metrics, both traditional and learning-based, have been proposed, making it a challenge to select the appropriate method for a given task. This study presents a comparative analysis between five widely used traditional no-reference image quality assessment techniques and five machine learning-based approaches, evaluating their effectiveness in computing image quality scores. The evaluation is carried out comprehensively using a set of standard and advanced performance metrics. Furthermore, we analyze how characteristics of the training datasets, such as score distribution, influence model performance. The machine learning models considered vary significantly in architectural complexity, in terms of both the number of layers and parameters, and we investigate whether this variability has a considerable impact on prediction accuracy. The analysis also extends to non-photographic imagery, with a comparative evaluation of the methods on hyperspectral satellite image visualizations. For full transparency and reproducibility of the current study, all training parameters and hardware specifications are reported.

Enhancing Domain-Specific Named Entity Recognition via Segmentation and Pseudo-Labeled Annotation (accepted at ICTAI 2025)
🧑 Pape Ibrahima Thiam, Yohann Chasseray, Josiane Mothe, Mathieu Roche, and Maguelonne Teisseire
📅 November 2025
Accepted as full paper in the Proceedings of the 37th IEEE International Conference on Tools with Artificial Intelligence (ICTAI 2025).
This year ICTAI received a record 633 submissions in total. Of these, only 133 were accepted as full papers (21.0% acceptance rate)
Territorial food systems, captured here through French-language documents, demand robust Named Entity Recognition (NER) despite scarce annotations and long, heterogeneous texts. In this work, we tackle two challenges: (1) segmenting long documents and (2) adapting open-schema NER to a specialized, low-resource domain. First, we compare four segmentation strategies in zero-shot settings to quantify precision–recall trade-offs. Then, we propose a semi-supervised pipeline that fine-tunes GLiNER and NuNER using a small manually annotated seed set followed by a large pseudo-labeled corpus built via cross-model agreement. Evaluations on a test set and the full annotated corpus show that pseudo-label fine-tuning consistently outperforms training on human-labeled data alone. The study also exposes strategy-specific strengths and weaknesses, underscoring that optimizing segmentation materially affects NER in domain-specific, low-resource scenarios. Our results provide practical guidance for deploying NER in territorial food systems and comparable specialized domains.
Publishing managers: J. Mothe & S. Molina, UT3 & UT2, IRIT, France