Table 6.

Classification results of four deep learning classifiers on GasHisSDB [In (%)].

Sub-database size	Model	Quantity of epoch	Model size (MB)	Best eopch	Training time(s)	Acc	Category	Pre	Rec	Spe	F1
160 × 160 pixels	VGG16	100	268.16	100	13,873	95.90	Abnormal	93.8	96.0	95.9	94.9
							Normal	97.3	95.9	96.0	96.6
	Inception-V3	100	89.69	92	10,296	94.57	Abnormal	94.1	92.0	96.2	93.0
							Normal	94.9	96.2	92.0	95.5
	ResNet50	100	83.12	84	10,023	96.09	Abnormal	94.6	95.6	96.4	95.1
							Normal	97.1	96.4	95.6	96.7
	ViT	100	31.17	97	2,587	86.21	Abnormal	83.8	80.6	89.9	82.2
							Normal	87.7	89.9	80.6	88.8
		400	31.17	399	10,014	92.23	Abnormal	92.1	87.8	95.1	89.9
							Normal	92.3	95.1	87.8	93.7
120 × 120 pixels	VGG16	100	268.16	100	26,105	96.47	Abnormal	96.7	94.0	98.0	95.3
							Normal	96.4	98.0	94.0	97.2
	Inception-V3	100	89.69	98	19,719	95.83	Abnormal	94.6	94.4	96.7	94.5
							Normal	96.6	96.7	94.4	96.6
	ResNet50	100	83.12	94	19,087	95.94	Abnormal	96.2	93.0	97.8	94.6
							Normal	95.8	97.8	93.0	96.8
	ViT	100	31.17	100	4,077	89.44	Abnormal	87.0	84.9	92.2	85.9
							Normal	90.9	92.2	84.9	91.5
		500	31.17	496	20,410	94.59	Abnormal	93.5	93.4	95.3	93.2
							Normal	95.4	95.9	92.5	95.6
80 × 80 pixels	VGG16	100	268.16	90	62,152	96.12	Abnormal	94.2	96.3	96.0	95.2
							Normal	97.4	96.0	96.3	96.7
	Inception-V3	100	89.69	99	43,926	95.41	Abnormal	95.5	93.0	97.0	94.2
							Normal	95.3	97.0	93.0	96.1
	ResNet50	100	83.12	97	41,992	96.09	Abnormal	96.2	94.0	97.5	95.1
							Normal	96.0	97.5	94.0	96.7
	ViT	100	31.17	89	8,247	90.23	Abnormal	86.3	90.1	90.3	88.2
							Normal	93.1	90.3	90.1	91.7
		500	31.17	496	41,135	94.57	Abnormal	93.1	93.4	95.3	93.2
							Normal	95.6	95.3	93.4	95.4

The bold text in the table indicates the maximum value or the best index of the classification results of different categories.