On the Effect of Pre-Processing and Model Complexity for Plastic Analysis Using Short-Wave-Infrared Hyper-Spectral Imaging

TL;DR

This paper conducts an extensive empirical study on hyper-spectral image segmentation of plastics, emphasizing model complexity, pre-processing, and introducing a new specialized architecture and dataset for improved recycling analysis.

Contribution

It introduces PlasticNet, a specialized hyper-spectral segmentation model, and provides a comprehensive dataset, highlighting the importance of model simplicity and pre-processing in plastic analysis.

Findings

Specialized models outperform generic ones in accuracy and efficiency.

Pre-processing significantly improves segmentation performance.

PlasticNet variants outperform other architectures in both accuracy and computational cost.

Abstract

The importance of plastic waste recycling is undeniable. In this respect, computer vision and deep learning enable solutions through the automated analysis of short-wave-infrared hyper-spectral images of plastics. In this paper, we offer an exhaustive empirical study to show the importance of efficient model selection for resolving the task of hyper-spectral image segmentation of various plastic flakes using deep learning. We assess the complexity level of generic and specialized models and infer their performance capacity: generic models are often unnecessarily complex. We introduce two variants of a specialized hyper-spectral architecture, PlasticNet, that outperforms several well-known segmentation architectures in both performance as well as computational complexity. In addition, we shed lights on the significance of signal pre-processing within the realm of hyper-spectral imaging. To complete our contribution, we introduce the largest, most versatile hyper-spectral dataset of plastic flakes of four primary polymer types.