Leonardo vindicated: Pythagorean trees for minimal reconstruction of the natural branching structures

Dymitr Ruta,Corrado Mio,Ernesto Damiani

2024-11-13

Abstract:Trees continue to fascinate with their natural beauty and as engineering masterpieces optimal with respect to several independent criteria. Pythagorean tree is a well-known fractal design that realistically mimics the natural tree branching structures. We study various types of Pythagorean-like fractal trees with different shapes of the base, branching angles and relaxed scales in an attempt to identify and explain which variants are the closest match to the branching structures commonly observed in the natural world. Pursuing simultaneously the realism and minimalism of the fractal tree model, we have developed a flexibly parameterised and fast algorithm to grow and visually examine deep Pythagorean-inspired fractal trees with the capability to orderly over- or underestimate the Leonardo da Vinci's tree branching rule as well as control various imbalances and branching angles. We tested the realism of the generated fractal tree images by means of the classification accuracy of detecting natural tree with the transfer-trained deep Convolutional Neural Networks (CNNs). Having empirically established the parameters of the fractal trees that maximize the CNN's natural tree class classification accuracy we have translated them back to the scales and angles of branches and came to the interesting conclusions that support the da Vinci branching rule and golden ratio based scaling for both the shape of the branch and imbalance between the child branches, and claim the flexibly parameterized fractal trees can be used to generate artificial examples to train robust detectors of different species of trees.

Artificial Intelligence,Machine Learning

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: how to most realistically reconstruct the branching structure of natural trees through the improved fractal tree model (especially the Pythagorean tree), and verify the similarity between these models and real trees.

Specifically, the researchers focus on the following aspects:

1. **Modeling the Branching Structure of Natural Trees**:
- Research different types of Pythagorean trees and their variants. By adjusting the basic shape, branching angle, and scale, explore which variants are closest to the common tree branching structures in nature.
- Use a fast algorithm with flexible parameterization to generate and visually inspect deep Pythagorean - inspired fractal trees, which can orderly exceed or underestimate the Leonardo da Vinci tree branching rules and control various imbalances and branching angles.

2. **Verifying the Fidelity of the Model**:
- Classify the generated fractal tree images through a well - trained deep convolutional neural network (CNN) to evaluate its ability to recognize real trees. The specific method is to use transfer learning technology to fine - tune the pre - trained GoogLeNet model on a dataset containing a large number of real - tree images.
- Determine which parameter combinations of the fractal tree are closest to the real trees according to the classification accuracy of the CNN.

3. **Supporting Leonardo da Vinci Branching Rules and the Golden Ratio**:
- The research finds that the fractal tree models that support Leonardo da Vinci branching rules (that is, the sum of the cross - sectional areas of all branches is equal to the cross - sectional area of the trunk) and the scaling factor based on the golden ratio (used to describe the imbalance between branch shapes and sub - branches) can more realistically simulate the branching structure of natural trees.
- These models are not only visually closer to real trees but also show higher accuracy in the CNN classification test.

4. **Generating Artificial Data to Train Robust Detectors**:
- Propose that artificial examples can be generated using flexibly parameterized fractal trees to train robust detectors that can recognize different tree species.

In summary, this paper aims to find a fractal tree model that can both remain simple and highly realistically simulate the branching structure of natural trees through mathematical and machine - learning methods, and verify its effectiveness through experiments.