Automatic identifying and counting blood cells in smear images by using single shot detector and Taguchi method

BMC Bioinformatics

Table 4 mAP, average mAP, SD, and η value achieved by the Resnet50 backbone network of the SSD in detecting different blood cells in smear images when the combinations of algorithm hyperparameters in Table 3 were employed in three independent experimental runs

Experiment no	Dataset	mAP-Experiment no			Average mAP	SD	η value
Experiment no	Dataset	mAP-1	mAP-2	mAP-3	Average mAP	SD	η value
1	Training set	0.7322	0.7194	0.7111	0.7209	0.0106	11.0848
1	Test set	0.7179	0.7113	0.6713	0.7002	0.0252	10.4624
2	Training set	0.7715	0.772	0.7656	0.7697	0.0036	12.7541
2	Test set	0.7705	0.7679	0.7672	0.7685	0.0017	12.7102
3	Training set	0.6245	0.6242	0.6247	0.6245	0.0003	8.5070
3	Test set	0.6852	0.6853	0.6851	0.6852	0.0001	10.0393
4	Training set	0.1907	0.1902	0.1885	0.1898	0.0012	1.8282
4	Test set	0.19	0.1896	0.1874	0.1890	0.0014	1.8196
5	Training set	0.0332	0.0332	0.0332	0.0332	0.0000	0.2933
5	Test set	0.0373	0.0374	0.0372	0.0373	0.0001	0.3302
6	Training set	0.4613	0.4555	0.455	0.4573	0.0035	5.3083
6	Test set	0.4729	0.472	0.4716	0.4722	0.0007	5.5501
7	Training set	0.6348	0.6332	0.6338	0.6339	0.0008	8.7288
7	Test set	0.6961	0.695	0.6938	0.6950	0.0012	10.3131
8	Training set	0.7082	0.7161	0.733	0.7191	0.0127	11.0290
8	Test set	0.6913	0.719	0.7311	0.7138	0.0204	10.8666
9	Training set	0.7767	0.7797	0.7821	0.7795	0.0027	13.1318
9	Test set	0.7653	0.7646	0.7686	0.7662	0.0021	12.6219

ISSN: 1471-2105