ParticleSeg3D: A Scalable Out-of-the-Box Deep Learning Segmentation Solution for Individual Particle Characterization from Micro CT Images in Mineral Processing and Recycling
Authors:
Karol Gotkowski,
Shuvam Gupta,
Jose R. A. Godinho,
Camila G. S. Tochtrop,
Klaus H. Maier-Hein,
Fabian Isensee
Abstract:
Minerals, metals, and plastics are indispensable for a functioning modern society. Yet, their supply is limited causing a need for optimizing ore extraction and recuperation from recyclable materials.Typically, those processes must be meticulously adapted to the precise properties of the processed materials. Advancing our understanding of these materials is thus vital and can be achieved by crushi…
▽ More
Minerals, metals, and plastics are indispensable for a functioning modern society. Yet, their supply is limited causing a need for optimizing ore extraction and recuperation from recyclable materials.Typically, those processes must be meticulously adapted to the precise properties of the processed materials. Advancing our understanding of these materials is thus vital and can be achieved by crushing them into particles of micrometer size followed by their characterization. Current imaging approaches perform this analysis based on segmentation and characterization of particles imaged with computed tomography (CT), and rely on rudimentary postprocessing techniques to separate touching particles. However, their inability to reliably perform this separation as well as the need to retrain methods for each new image, these approaches leave untapped potential to be leveraged. Here, we propose ParticleSeg3D, an instance segmentation method able to extract individual particles from large CT images of particle samples containing different materials. Our approach is based on the powerful nnU-Net framework, introduces a particle size normalization, uses a border-core representation to enable instance segmentation, and is trained with a large dataset containing particles of numerous different sizes, shapes, and compositions of various materials. We demonstrate that ParticleSeg3D can be applied out-of-the-box to a large variety of particle types, including materials and appearances that have not been part of the training set. Thus, no further manual annotations and retraining are required when applying the method to new particle samples, enabling substantially higher scalability of experiments than existing methods. Our code and dataset are made publicly available.
△ Less
Submitted 14 December, 2023; v1 submitted 30 January, 2023;
originally announced January 2023.
