TABLE 5.

Performance comparison of the presented framework with machine learning (ML)-based classifiers.

Classifier	Execution time (s)	Total trainable model parameters (Million)	Accuracy
Deep-keypoints along with the RF classifier (Ma et al., 2018)	–	–	93.4%
Deep-keypoints along with the ELM classifier (Xian and Ngadiran, 2021)	–	–	84.94%
Deep-keypoints along with the DT classifier (Xian and Ngadiran, 2021)	–	–	77.8%
Deep-keypoints along with the SVM classifier (Mohameth et al., 2020)	12 mn 21	25.5	98.01%
Deep-keypoints along with the KNN classifier (Mohameth et al., 2020)	12 mn 21	25.5	91.01%
Proposed	17.5	14.4	99.99%

Bold means the architectures are improved.