CASE STUDY 08
- Academic
- Machine Learning
Predicting plant diversity
Multi-label prediction from multimodal data.
GeoLifeCLEF 2025 · Personal / academic project
The project, step by step
Multi-labelplant species prediction
- 01
Multimodal data
GeoLifeCLEF 2025
- 02
Train the model
TabularMLP pipeline on a GPU server
- 03
Predict species
Multiple possible species per observation
- 04
Analyse results
Modality ablations and explainability
Context
Plant species prediction project as part of GeoLifeCLEF 2025.
Problem
Predict multiple plant species from multimodal data and study the contribution of those modalities.
Data
Multimodal data from GeoLifeCLEF 2025. The data volumes and modalities used remain to be specified.
Approach
I developed a TabularMLP pipeline, trained it on a shared GPU server and carried out modality ablations.
Method
Pipeline training, ablations and explainability analysis with Grad-CAM and SHAP. The components addressed by each method remain to be specified.
Results
A TabularMLP pipeline developed, with explainability and ablation analyses completed. The score and ranking are not provided.
Technologies
PyTorch, Scikit-learn, Grad-CAM and SHAP.
Lessons and limitations
The contribution of each modality, ablation findings and model limitations remain to be documented.