A Sustainable Deep Learning Framework for Object Recognition Using Multi-Layers Deep Features Fusion and Selection
Published in Sustainability (MDPI), 2020
Authors:
Muhammad Rashid, Muhammad Attique Khan, Majed Alhaisoni, Shui-Hua Wang, Syed Rameez Naqvi, Amjad Rehman, Tanzila Saba
This work presents a sustainable deep learning framework for object recognition using multi-layer deep feature extraction, fusion, and selection. The method uses transfer learning with VGG19 and Inception V3, extracts deep features from selected layers, fuses them using a Parallel Maximum Covariance (PMC) approach, and selects the most discriminative features using a Multi Logistic Regression controlled Entropy-Variances (MRcEV) method.
The framework was evaluated on four public datasets: Caltech-101, Birds, Butterflies, and CIFAR-100. The reported accuracies were 95.5%, 100%, 98%, and 68.80%, respectively.
Contributions 📃
- A sustainable deep learning framework for complex object classification.
- Transfer-learning-based feature extraction using VGG19 and Inception V3.
- Feature fusion through Parallel Maximum Covariance (PMC).
- Robust feature selection using MRcEV.
- Evaluation on multiple public object recognition datasets.
Dataset and Model Summary
| Component | Details |
|---|---|
| Task | Object recognition / object classification |
| Models | VGG19, Inception V3 |
| Feature Fusion | Parallel Maximum Covariance |
| Feature Selection | MRcEV |
| Classifier | Ensemble Subspace Discriminant |
| Datasets | Caltech-101, Birds, Butterflies, CIFAR-100 |
| Journal | Sustainability |
| Publisher | MDPI |
| Publication Date | 19 June 2020 |
📖 Citation (BibTeX)
@article{rashid2020sustainable,
title = {A Sustainable Deep Learning Framework for Object Recognition Using Multi-Layers Deep Features Fusion and Selection},
author = {Rashid, Muhammad and Khan, Muhammad Attique and Alhaisoni, Majed and Wang, Shui-Hua and Naqvi, Syed Rameez and Rehman, Amjad and Saba, Tanzila},
journal = {Sustainability},
volume = {12},
number = {12},
pages = {5037},
year = {2020},
publisher = {MDPI},
doi = {10.3390/su12125037}
}