Fig 1.
Sample images from the four distinct waste material datasets.
Table 1.
An overview of the datasets.
Table 2.
Unique waste classes throughout the 7 datasets.
Fig 2.
Customized DenseNet-201 with Squeeze and Excitation Integrated Parallel CNN Architecture.
Fig 3.
Squeeze and Excitation Mechanism.
Table 3.
Experimental setup for the proposed classification model.
Fig 4.
Accuracy & loss curves of the DenseNet201-Infused Parallel SE-CNN for Waste Classification V2 dataset.
Fig 5.
Confusion matrix for Waste Classification V2 dataset.
Table 4.
Classwise metrics overview for Waste classification V2 dataset.
Fig 6.
Overall Metrics overview for Waste Classification V2 dataset.
Fig 7.
ROC curve for Waste Classification V2 dataset.
Fig 8.
Accuracy & loss curves of the DenseNet201-Infused Parallel SE-CNN for Waste Classification dataset.
Fig 9.
Confusion Matrix for Waste Classification dataset.
Table 5.
Classwise metrics overview for the waste classification dataset.
Fig 10.
Overall Metrics overview for Waste Classification dataset.
Fig 11.
ROC Curve for Waste Classification dataset.
Fig 12.
Accuracy & loss curves of the DenseNet201-Infused Parallel SE-CNN for OpenRecycle Dataset.
Fig 13.
Confusion Matrix for OpenRecycle Dataset.
Table 6.
Classwise metrics overview for the OpenRecycle dataset.
Fig 14.
Overall Metrics overview for the OpenRecycle dataset.
Fig 15.
ROC Curve for OpenRecycle Dataset.
Fig 16.
Accuracy & loss curves of the DenseNet201-Infused Parallel SE-CNN for TrashNet Dataset.
Fig 17.
Confusion Matrix for TrashNet Dataset.
Table 7.
Classwise metrics overview for the TrashNet dataset.
Table 8.
Classwise metrics overview after data augmentation for TrashNet dataset.
Fig 18.
Overall Metrics overview for TrashNet dataset.
Fig 19.
ROC Curve for TrashNet Dataset.
Table 9.
Robustness evaluation of model under adversarial attacks, noisy data, and environmental conditions.
Table 10.
Performance metrics of the model across three additional datasets.
Table 11.
Comparative analysis of classification performance with different deep CNN models.
Table 12.
ML classifiers performance comparison across datasets.
Table 13.
Comparative analysis of classification performance among models from prior studies.
Table 14.
Impact of parallel CNN Branches and SE attention on model performance.
Table 15.
Empirical time complexity analysis across datasets.
Table 16.
Energy consumption and carbon emissions for different datasets.
Table 17.
Empirical benchmarking of model complexity and computational requirements across architectures.
Fig 20.
Explainability of Proposed Model Using Grad-CAM.
Fig 21.
Interface of the Website.
Fig 22.
Uploading an image and predicting via website.
Fig 23.
Gradio Deployment.