Vision-Based Deep Q-Learning Network Models to Predict Particulate Matter Concentration Levels Using Temporal Digital Image Data

Table 5

Prediction accuracy of predictive models with an application to various video sequences.

# of frames	5 frames	10 frames	15 frames	20 frames
	Konkuk Univ. Indoor
Deep Q-Haze	0.9918	0.9945	0.9927	0.9817
Deep Haze	0.7112	0.7487	0.7450	0.7487
RF	0.6262	0.6300	0.5000	0.5000
SVM	0.5000	0.5000	0.5000	0.5000
	Konkuk Univ. Outdoor
Deep Q-Haze	0.9040	0.9100	0.9160	0.9220
Deep Haze	0.4560	0.4580	0.4520	0.4580
RF	0.4690	0.5060	0.4380	0.4610
SVM	0.6560	0.6270	0.6240	0.6440
	Keimyung Univ.
Deep Q-Haze	0.9839	0.9861	0.9877	0.9914
Deep Haze	0.6300	0.6336	0.6309	0.6336
RF	0.4100	0.4390	0.4418	0.4581
SVM	0.3800	0.4054	0.4318	0.4236
	Mobile Phone
Deep Q-Haze	0.8842	0.8796	0.8703	0.8657
Deep Haze	0.5046	0.5000	0.5023	0.5000
RF	0.4768	0.5046	0.4513	0.4791
SVM	0.3287	0.3217	0.3148	0.3148