Application of Deep Learning in Integrated Pest Management: A Real-Time System for Detection and Diagnosis of Oilseed Rape Pests

Table 4

The results for five different architectures.

Meta-architecture	Faster R-CNN	R-FCN		SSD
Feature extractor	ResNet101	Inception	ResNet101	Inception	MobileNet
Athalia rosae japanensis	0.8321	0.8393	0.7703	0.8218	0.8302
Creatonotus transiens	0.5743	0.5527	0.6083	0.586	0.5951
Entomoscelis adonidis	0.8765	0.8246	0.9161	0.6176	0.6030
Entomoscelis suturalis	0.6562	0.4162	0.4344	0.7555	0.4980
Hellula undalis	0.7563	0.7029	0.7106	0.7738	0.6247
Lipaphis erysimi	0.2149	0.3003	0.3606	0.3772	0.3236
Mamestra brassicae	0.9288	0.8445	0.9280	0.967	0.8431
Meligethes aeneus	0.5375	0.2637	0.3556	0.2247	0.3476
Phyllotreta striolata	0.7604	0.5197	0.6124	0.6275	0.6760
Pieris rapae	0.8143	0.8040	0.8432	0.9111	0.6609
Plutella xylostella	0.8629	0.7890	0.7880	0.8767	0.7846
Psylliodes punctifrons	0.4537	0.4202	0.5368	0.554	0.7593
[email protected]	0.6890	0.6064	0.6554	0.6744	0.6288
Time (s)	0.158	0.13	0.148	0.052	0.045
Memory (MB)	191.3	52.9	201.7	60.6	23.6