MiniFarm | Synthetic Botany Research

Computational Botany Frameworks

Evaluation of high-tier software for parametrically generating 3D plant models. The objective is to identify the optimal engine for creating MiniFarm's synthetic training datasets, balancing biological fidelity with computer vision export capabilities.

The Trade-off Landscape

Our analysis reveals a distinct dichotomy in the current market. FSPM (Functional-Structural Plant Models) excel at biological correctness but struggle with photo-realism and scale. Conversely, Game Engines & Synthetic Data generators offer superior rendering and ground-truth extraction but often lack deep botanical logic without significant custom plugin development.

FSPM: High Bio Accuracy Engines: High Rendering DCC: High Control

Key Insight

For the MiniFarm project, a hybrid approach is recommended: utilizing FSPM logic (e.g., L-Py) to generate topology, then exporting to a Synthetic Engine (e.g., Omniverse) for texturing, disease layering, and rendering.

Category Capability Profiles

Best for Biology

OpenAlea / L-Py

Unmatched physiological simulation capability.

Best for AI Data

NVIDIA Omniverse

Native semantic segmentation & USD workflows.

Best for Complexity

GroIMP

Handles massive environmental interaction graphs.