[Retracted] Multipurpose Watermarking Approach for Copyright and Integrity of Steganographic Autoencoder Models

Table 2

Fidelity evaluation of the three sets of our watermarking model.

Structure–Dataset	Class	Baseline accuracy (%)	Rate of watermarked weights (%)	Watermarked accuracy (%)
Coco-2014	All	96.02	15	95.51
	Cat	96.04	15	95.54
	Car	96.00	15	95.51
	Banana	95.94	15	95.51
	Person	95.88	15	95.43
Coco-2017	All	96.10	15	95.58
	Cat	96.16	15	95.52
	Car	96.00	15	95.47
	Banana	96.07	15	95.56
	Person	96.14	15	95.57
Boss	All	96.05	15	95.52