An optimised YOLOv4 deep learning model for efficient malarial cell detection in thin blood smear images

Parasites & Vectors

Table 5 Network architecture of the pruned models

Candidate model	Res-block 3 (C3)	Res-block 4 (C4)	Res-block 5 (C5)
Candidate model	1 × 1/1 3 × 3/1
Candidate 1 (YOLOv4-RC3)	'x' [52 × 52 × 128]	‘/’ [26 × 26 256]	‘/’ [13 × 13 × 512]
Candidate 2 (YOLOv4-RC4)	‘/’ [52 × 52 × 128]	'x' [26 × 26 256]	‘/’ [13 × 13 × 512]
Candidate 3 (YOLOv4-RC5)	‘/’ [52 × 52 × 128]	‘/’ [52 × 52 × 128]	'x' [13 × 13 × 512]
Candidate 4 (YOLOv4-RC3_4)	'x' [52 × 52 × 128]	'x' [26 × 26 256]	‘/’ [13 × 13 × 512]
Candidate 5 (YOLOv4-RC3_5)	'x' [52 × 52 × 128]	‘/’ [26 × 26 256]	'x' [13 × 13 × 512]

'x' indicates that pruning occurs on the Res-Block bodies; ‘/’ indicates that no pruning occurs on the Res-block bodies; values in square brackets indicate the image output sizes on the remaining layers of the Res-block bodies
YOLO You Only Look Once (model)

ISSN: 1756-3305