基于机器视觉和机器学习技术的浙贝母外观品质等级区分

董成烨,李东方,冯槐区,龙思放,奚特,周芩安,王俊

Classification of Fritillaria thunbergii appearance quality based on machine vision and machine learning technology

Chengye DONG,Dongfang LI,Huaiqu FENG,Sifang LONG,Te XI,Qin’an ZHOU,Jun WANG

表 3 不同空洞率的YOLOX-DC训练所得模型在浙贝母测试集上的测试结果

Table 3 Test results of models trained by YOLOX-DC with different dilated rates on the test set

空洞率 Dilated rate	指标 Index	浙贝母品质 Quality of F. thunbergii
空洞率 Dilated rate	指标 Index	特级 Superfine	一级 Level one	二级 Level two	虫蛀 Moth-eaten	霉变 Mildewed	破碎 Broken
2	AP/%	99.51	93.53	96.97	97.85	99.77	97.48
	F₁	0.95	0.84	0.81	0.96	0.98	0.97
	mAP/%	97.52
	FPS	28.97
3	AP/%	99.37	98.52	97.62	97.75	99.93	98.37
	F₁	0.97	0.94	0.93	0.97	0.99	0.95
	mAP/%	98.59
	FPS	29.18
4	AP/%	99.97	98.33	98.47	98.71	99.73	98.85
	F₁	0.99	0.92	0.94	0.97	0.99	0.97
	mAP/%	99.01
	FPS	29.13
5	AP/%	98.63	95.83	94.89	98.34	99.86	98.99
	F₁	0.94	0.81	0.89	0.96	0.98	0.97
	mAP/%	97.76
	FPS	29.02
6	AP/%	98.16	94.31	98.77	98.65	99.55	97.11
	F₁	0.92	0.86	0.75	0.97	0.92	0.96
	mAP/%	97.76
	FPS	29.05