An Improved Self-Training Model with Fine-Tuning Teacher/Student Exchange Strategy to Detect Computer-Generated Images

Table 11

Test results for different dataset-combination settings with attacks to 50% of the testing dataset.

Model	Accuracy (%)			Precision (%)			Recall (%)
	Comb1	Comb2	Comb3	Comb1	Comb2	Comb3	Comb1	Comb2	Comb3
M0	71.93	75.44	82.13	86.40	77.44	96.54	66.99	74.45	74.93
M1	84.40	82.65	91.73	91.00	83.56	94.83	80.38	82.05	89.28
M2	84.85	85.04	91.25	90.09	86.44	94.96	81.53	84.08	88.39
M3	85.12	84.22	91.64	90.84	83.85	94.60	81.51	84.46	89.30