Table 11.

The performance of DenseNet201 with incorporated channel attention on different dense blocks for classification

CNN Models	Batch Size	Precision	Recall	F1-score
DenseNet201	16	89%	87%	88%
DenseNet201	32	87%	84%	85%
DenseNet201	64	82%	83%	82%
DenseNet201+ CA + SA	92	88%	90%	67%
DenseNet201+ CA on $4^{\mathit{th}}$ Block	16	91%	88%	89%
DenseNet201+ CA on $3^{\mathit{rd}}$ Block	16	91%	88%	89%
DenseNet201+ attention on $3^{\mathit{rd}}$ Block and $4^{\mathit{th}}$ block	16	94%	93%	93%
Proposed model without image enhancement	16	95%	96%	95%
Proposed model with CLAHE-enhanced image	16	97%	96%	96%