Page 154 - 2023-Vol19-Issue2

150 |
Fig. 2. Proposed EL-based CBIR system
C. Xception Model
It is the Inception architecture in a more developed form. A
linear stack of depth-wise separable convolution layers with
lingering connections is what it is, according to [34]. These
layers aid in lowering the need for memory and the expense of
computing. The 14 modules of the 36 convolutional layers that
make up Xception all feature linear residual connections, with
the exception of the first and final modules. By dividing the
separable convolution in Xception, space-wise and channel-
wise features are learned.
D. MobileNet Model
The core of the MobileNet model is depth wise separable con-
volutions, which factorize a standard convolution into a depth
wise convolution and an additional convolution known as a
pointwise convolution. Each input channel is subjected to a
single filter during the depth wise convolution for MobileNets. ception, and MobileNet supervised learning algorithms were
The pointwise convolution employing an 11 convolution then
combines the results of the depth wise convolution. Standard
Mohammed, Oraibi & Hussain
convolutions filter the inputs and combine them into a new set
of outputs in one step. The depth wise separable convolution
separates this into two layers: one for mixing and filtering,
and another layer [35].
E. Ensemble Hard Voting (HV) Model
An example of a voting algorithm is a meta-classifier that
assembles similar or conceptually different ML classifiers for
prediction via voting. It serves as a container for a collection
of several classifiers that have been simultaneously trained
and assessed to take use of the unique characteristics of each
method. A voting method with less overfitting and less in
accuracy is HV, which is the simplest instance. According to
the variation classifiers, HV will be the most common class la-
bel [36]. On several image datasets, an HV meta-classifier has
been used for the final classification stage. The Xception, In-
used to create the HV meta-classifier. To increase forecast
accuracy, ensemble voting may be crucial.