Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neuroplastic Modular Framework: Cross-Domain Image Classification of Garbage and Industrial Surfaces

Debojyoti Ghosh, Soumya K Ghosh, Adrijit Goswami

TL;DR

The paper tackles robust image classification for dynamic waste streams and industrial surface defects by proposing the Neuroplastic Modular Classifier, a hybrid system that fuses ResNet-50 local features, Vision Transformer global context, and FAISS-based memory retrieval, augmented by dynamically growing modular blocks inspired by neuroplasticity. It introduces a memory-augmented fusion layer and a self-expanding architecture that adds capacity when performance stagnates, enabling continual adaptation across domains. Empirical results show state-of-the-art accuracy on garbage classification (97.15%) and near-perfect industrial defect detection on KolektorSDD2 (99.80%), supported by ablation and statistical significance analyses. The work demonstrates strong cross-domain generalization, scalability, and practical deployment potential for smart waste management and automated industrial inspection, with future directions toward continual learning and meta-learning-driven growth optimization.

Abstract

Efficient and accurate classification of waste and industrial surface defects is essential for ensuring sustainable waste management and maintaining high standards in quality control. This paper introduces the Neuroplastic Modular Classifier, a novel hybrid architecture designed for robust and adaptive image classification in dynamic environments. The model combines a ResNet-50 backbone for localized feature extraction with a Vision Transformer (ViT) to capture global semantic context. Additionally, FAISS-based similarity retrieval is incorporated to provide a memory-like reference to previously encountered data, enriching the model's feature space. A key innovation of our architecture is the neuroplastic modular design composed of expandable, learnable blocks that dynamically grow during training when performance plateaus. Inspired by biological learning systems, this mechanism allows the model to adapt to data complexity over time, improving generalization. Beyond garbage classification, we validate the model on the Kolektor Surface Defect Dataset 2 (KolektorSDD2), which involves industrial defect detection on metal surfaces. Experimental results across domains show that the proposed architecture outperforms traditional static models in both accuracy and adaptability. The Neuroplastic Modular Classifier offers a scalable, high-performance solution for real-world image classification, with strong applicability in both environmental and industrial domains.

Neuroplastic Modular Framework: Cross-Domain Image Classification of Garbage and Industrial Surfaces

TL;DR

The paper tackles robust image classification for dynamic waste streams and industrial surface defects by proposing the Neuroplastic Modular Classifier, a hybrid system that fuses ResNet-50 local features, Vision Transformer global context, and FAISS-based memory retrieval, augmented by dynamically growing modular blocks inspired by neuroplasticity. It introduces a memory-augmented fusion layer and a self-expanding architecture that adds capacity when performance stagnates, enabling continual adaptation across domains. Empirical results show state-of-the-art accuracy on garbage classification (97.15%) and near-perfect industrial defect detection on KolektorSDD2 (99.80%), supported by ablation and statistical significance analyses. The work demonstrates strong cross-domain generalization, scalability, and practical deployment potential for smart waste management and automated industrial inspection, with future directions toward continual learning and meta-learning-driven growth optimization.

Abstract

Efficient and accurate classification of waste and industrial surface defects is essential for ensuring sustainable waste management and maintaining high standards in quality control. This paper introduces the Neuroplastic Modular Classifier, a novel hybrid architecture designed for robust and adaptive image classification in dynamic environments. The model combines a ResNet-50 backbone for localized feature extraction with a Vision Transformer (ViT) to capture global semantic context. Additionally, FAISS-based similarity retrieval is incorporated to provide a memory-like reference to previously encountered data, enriching the model's feature space. A key innovation of our architecture is the neuroplastic modular design composed of expandable, learnable blocks that dynamically grow during training when performance plateaus. Inspired by biological learning systems, this mechanism allows the model to adapt to data complexity over time, improving generalization. Beyond garbage classification, we validate the model on the Kolektor Surface Defect Dataset 2 (KolektorSDD2), which involves industrial defect detection on metal surfaces. Experimental results across domains show that the proposed architecture outperforms traditional static models in both accuracy and adaptability. The Neuroplastic Modular Classifier offers a scalable, high-performance solution for real-world image classification, with strong applicability in both environmental and industrial domains.

Paper Structure

This paper contains 40 sections, 23 equations, 7 figures, 8 tables, 4 algorithms.

Table of Contents

  1. Introduction
  2. Literature Survey
  3. Deep CNN-Based Models
  4. Hybrid and Transfer Learning Models
  5. Optimization and Lightweight Architectures
  6. Emerging Trends in Hazard and Industrial Waste Management
  7. Domain-Inspired Modular Architectures
  8. Research Gap and Motivation
  9. Contribution
  10. Dataset Overview and Preprocessing Workflow
  11. Model Architecture
  12. Feature Representation and Extraction
  13. Feature Fusion Layer
  14. Fusion Mechanism.
  15. Adaptive Neuroplastic Classification Network
  16. ...and 25 more sections

Figures (7)

  • Figure 1: Flowchart of the Neuroplastic Modular Classification pipeline combining ResNet, ViT, and FAISS-based memory into a dynamically growing architecture.
  • Figure 2: Flowchart of the feature fusion pipeline integrating ResNet, ViT, and FAISS features into a unified representation.
  • Figure 3: Flowchart of the Adaptive Modular Classifier
  • Figure 4: Flowchart of the neuroplastic module expansion logic.
  • Figure 5: Training pipeline with neuroplastic expansion and memory integration.
  • ...and 2 more figures