collective-similarity-index-quantitative-analysis

The Python scripts perform segmentation and analysis of microscopic images of recycled thermoplastic composite. The analysis is a 7-step pipeline that:

• Trains a Deep Learning model to convert microscopic images into binary images using EfficientNetB4 and PyTorch;

• Assembles images and performs segmentation of individual and multi-contact fibre footprints using ellipse-based methods;

• Performs a cluster analysis using Voronoi tessellation and computing the Collective Similarity Index (CSI) to identify fiber clusters;

• Computes microstructural descriptors such as Ripley's K function and pair distribution functions;

• Generates comprehensive visualizations of analysis results.

The code underlies the results published in Singh et al. (2026) [1]. The microscopic images of recycled thermoplastic composites analyzed in this work are publicly available via the 4TU.ResearchData archive (DOI: https://doi.org/10.4121/023f77ad-6ed9-4aa3-a88d-c5b2e09b7160) (see Related Dataset).

Related Dataset

D. Singh, A. Shakeel, and C. Dransfeld, 2026, Microscopic Images of Recycled Thermoplastic Composites. 4TU.ResearchData. Dataset. https://doi.org/10.4121/023f77ad-6ed9-4aa3-a88d-c5b2e09b7160