Table 3 Comparisons of detection performance and inference speed for YOLOV4-tiny by using the proposed and baseline approach on model development test set data
From: Tile-based microscopic image processing for malaria screening using a deep learning approach
Model | Tile size | AP (%) | R (%) | Time (sec/img) | |
|---|---|---|---|---|---|
Train | Inference | ||||
YOLOV4-tiny@416 | 1088 \(\times\) 1088 | 1088 \(\times\) 1088 | 56.0 | 60.0 | 1.0 |
832 \(\times\) 832 | 67.0 | 73.0 | 1.0 | ||
608 \(\times\) 608 | 78.0 | 86.0 | 1.4 | ||
416 \(\times\) 416 | 65.0 | 75.8 | 1.4 | ||
YOLOV4-tiny@512 | 1088 \(\times\) 1088 | 1088 \(\times\) 1088 | 70.0 | 75.9 | 1.0 |
832 \(\times\) 832 | 80.4 | 86.1 | 1.2 | ||
608 \(\times\) 608 | 82.6 | 91.4 | 1.6 | ||
YOLOV4-tiny@608 | 1088 \(\times\) 1088 | 1088 \(\times\) 1088 | 77.9 | 83.4 | 1.0 |
832 \(\times\) 832 | 85.9 | 92.7 | 1.3 | ||
608 \(\times\) 608 | 81.9 | 89.3 | 1.7 | ||
YOLOV4-tiny@416 | 832 \(\times\) 832 | 1088 \(\times\) 1088 | 53.3 | 57.6 | 1.0 |
832 \(\times\) 832 | 66.0 | 71.9 | 1.1 | ||
608 \(\times\) 608 | 81.4 | 88.1 | 1.6 | ||
YOLOV4-tiny@512 | 832 \(\times\) 832 | 1088 \(\times\) 1088 | 66.7 | 72.8 | 1.0 |
832 \(\times\) 832 | 79.1 | 85.0 | 1.1 | ||
608 \(\times\) 608 | 86.0 | 94.1 | 1.6 | ||
416 \(\times\) 416 | 78.8 | 89.2 | 2.7 | ||
YOLOV4-tiny@608 | 832 \(\times\) 832 | 1088 \(\times\) 1088 | 76.7 | 82.6 | 1.0 |
832 \(\times\) 832 | 86.0 | 92.9 | 1.3 | ||
608 \(\times\) 608 | 84.4 | 91.5 | 1.6 | ||
YOLOV4-tiny@416 | 608 \(\times\) 608 | 1088 \(\times\) 1088 | 48.0 | 57.1 | 1.0 |
832 \(\times\) 832 | 60.9 | 67.1 | 1.0 | ||
608 \(\times\) 608 | 83.2 | 90.1 | 1.3 | ||
416 \(\times\) 416 | 84.0 | 91.1 | 2.0 | ||
YOLOV4-tiny@512 | 608 \(\times\) 608 | 1088 \(\times\) 1088 | 54.3 | 60.8 | 1.0 |
832 \(\times\) 832 | 76.4 | 79.8 | 1.0 | ||
608 \(\times\) 608 | 87.1 | 95.3 | 1.5 | ||
416 \(\times\) 416 | 84.0 | 94.9 | 2.6 | ||
YOLOV4-tiny@608 | 608 \(\times\) 608 | 1088 \(\times\) 1088 | 69.4 | 85.7 | 1.0 |
832 \(\times\) 832 | 84.3 | 92.1 | 1.3 | ||
608 \(\times\) 608 | 87.0 | 95.3 | 1.9 | ||
Without tiling | |||||
YoloV4-tiny@416 | – | – | 54.0 | 21.0 | 0.15 |
YoloV4-tiny@512 | – | – | 69.0 | 48.0 | 0.15 |
YoloV4-tiny@608 | – | – | 76.0 | 57.0 | 0.15 |