Hybrid CNN-GNN architectures with distributed training for heathland plant classification
Babatoundé Moctard Olouladé, Jesper Leth Bak, Peter Borgen Sørensen, Haomin Yu, Christian Damgaard · 2026
Read Hybrid CNN-GNN architectures with distributed training for heathland plant classification by Babatoundé Moctard Olouladé, Jesper Leth Bak, Peter Borgen Sørensen, Haomin Yu, Christian Damgaard (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/pmc-13241450.txt.
Description
Olouladé and colleagues at Aarhus fused two neural network families to name heathland plants from the air. Standard image networks see pixels on grids, missing how plant parts relate spatially. Their PlantGraphNet turns image keypoints into graphs of connected plant regions. A convolutional backbone supplies appearance traits alongside.
On Danish aerial heathland data the hybrid reached near ninety-nine percent precision. That clearly beat image-only baselines on fine-grained species calls. Distributed training scaled almost linearly across rigs. Because decisions trace back to plant parts, the authors stress the system's interpretability for monitoring.
Catalog record
| CREATOR | Babatoundé Moctard Olouladé, Jesper Leth Bak, Peter Borgen Sørensen, Haomin Yu, Christian Damgaard |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 70 KB |
| VERSION | 1.0 |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |