Research Internship on "AI-Driven Field Boundary Detection: Leveraging Satellite Imagery to support Digital Agriculture adoption"

The Inria centre at Université Côte d'Azur includes 42 research teams and 9 support services. The centre's staff (about 500 people) is made up of scientists of different nationalities, engineers, technicians and administrative staff. The teams are mainly located on the university campuses of Sophia Antipolis and Nice as well as Montpellier, in close collaboration with research and higher education laboratories and establishments (Université Côte d'Azur, CNRS, INRAE, INSERM ...), but also with the regiona economic players.

With a presence in the fields of computational neuroscience and biology, data science and modeling, software engineering and certification, as well as collaborative robotics, the Inria Centre at Université Côte d'Azur  is a major player in terms of scientific excellence through its results and collaborations at both European and international levels.

The internship position is opened in the context of the EVERGREEN team project - https://team.inria.fr/evergreen/ .

Our team is actively working on the design and implementation of cutting-edge machine learning techniques to effectively exploit heterogeneous and multi-temporal Earth observation data for agricultrual and enviromental applications.

The team, located in a multidisciplinary laboratory, has an active and stimulating environment with master, PhD and Post-doc students coming from different countries.

Join our dynamic research team and you'll work on Transformer-based models to automatically extract agricultural field boundaries from satellite imagery [2].

What You'll Work On:

Depending on your interests, you'll explore one of two research tracks:

1) Transformer + Geometric Priors for Curvilinear Structures
Leverage the recent Time Series Vision Transformers (TsViT) model [1] in combination with procedural geometric priors, such as curvilinear structures [3], to accurately detect and model field boundaries—integrating both temporal and spatial insights to guide geometric extraction.

2) Foundation Models + Graph Neural Networks for End-to-End Vectorization
Combine a Semantic Segmentation Foundation Model (e.g., SAM2 [4]) with Graph Neural Networks to directly extract and vectorize agricultural fields in a single step—bridging raw data to precise geometric output [5,6].

You'll work with open-access, community-curated benchmark datasets, available via the public IEEE GRSS Data Portal: https://eod-grss-ieee.com/dataset-search

What You'll Gain:

• Hands-on experience with state-of-the-art AI models for geospatial analysis

• The opportunity to co-author a scientific publication, depending on your progress

• A collaborative and supportive environment at the forefront of AI for Earth observation
  

[1] Tarasiou, Michail, Erik Chavez, and Stefanos Zafeiriou. "Vits for sits: Vision transformers for satellite image time series." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2023.

[2] Kerner, Hannah, et al. "Fields of The World: A Machine Learning Benchmark Dataset For Global Agricultural Field Boundary Segmentation." arXiv preprint arXiv:2409.16252 (2024).

[3] Cheng, Mingfei, et al. "Joint topology-preserving and feature-refinement network for curvilinear structure segmentation." Proceedings of the IEEE/CVF International Conference on Computer Vision. 2021.

[4] Ravi, Nikhila, et al. "Sam 2: Segment anything in images and videos." arXiv preprint arXiv:2408.00714 (2024).

[5] Hetang, Congrui, et al. "Segment anything model for road network graph extraction." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2024.

[6] Adimoolam, Yeshwanth Kumar, Charalambos Poullis, and Melinos Averkiou. "Pix2poly: A sequence prediction method for end-to-end polygonal building footprint extraction from remote sensing imagery." 2025 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV). IEEE, 2025.

The selected intern will be expected to:

1. Review and analyze relevant scientific literature in deep learning and geospatial data processing.

2. Design a method to incorporate geometric prior knowledge into Transformer-based models or combine multiple models to extract and vectorize geometrical objects.

3. Implement the proposed approach using the PyTorch framework.

4. Conduct experiments and benchmark performance using real-world datasets for agricultural field boundary detection.

5. Document and communicate findings, including clear summaries, visualizations, and evaluations of the results.

 Technical Skills – Required Level

• Advanced Python programming

• Strong proficiency with data manipulation libraries (e.g., NumPy, Pandas)

• Expertise in deep learning frameworks such as PyTorch (preferred) or TensorFlow

• Experience with image analysis techniques, including segmentation, object detection, and classification

• Familiarity with satellite or remote sensing data is a plus

• Subsidized meals
• Partial reimbursement of public transport costs
• Leave: 7 weeks of annual leave + 10 extra days off due to RTT (statutory reduction in working hours) + possibility of exceptional leave (sick children, moving home, etc.)
• Possibility of teleworking and flexible organization of working hours
• Professional equipment available (videoconferencing, loan of computer equipment, etc.)
• Social, cultural and sports events and activities
• Access to vocational training
• Contribution to mutual insurance (subject to conditions)

Traineeship grant depending on attendance hours.