Visual Computing for Industry, Biomedicine, and Art

Table 9 Ablation study of attack effectiveness in terms of hyperparameters used in our method

From: Local imperceptible adversarial attacks against human pose estimation networks

	AP	AP⁵⁰	AP⁷⁵	AP^M	AP^L	AR	AR⁵⁰	AR⁷⁵	AR^M	AR^L
Clean	0.671	0.862	0.730	0.615	0.761	0.718	0.885	0.768	0.651	0.814
Ours (a = 1, b = 0)	0.605	0.802	0.658	0.516	0.739	0.653	0.826	0.696	0.552	0.796
Ours (a = 1, b = 0.5)	0.615	0.813	0.669	0.530	0.742	0.660	0.834	0.705	0.562	0.798
Ours (a = 1, b = 1)	0.639	0.833	0.698	0.562	0.761	0.692	0.866	0.741	0.606	0.814
Ours (a = 0.5, b = 1)	0.636	0.832	0.690	0.553	0.755	0.679	0.852	0.724	0.586	0.809
Ours (a = 0, b = 1)	0.642	0.834	0.700	0.563	0.759	0.686	0.857	0.732	0.596	0.811

Back to article page