Share your thoughts, 1 month free Claude Pro on usSee more

Adversarial Attack on ImageNet 1k 1000 images (Model Robustness)

100Robust Accuracy (Inc-v3)

BIM

Updated 1mo ago

Evaluation Results

Method	Links
BIM 2022.09		100	20.3	15.7	15.6	34.3	21.5
SIM 2022.09		100	38.2	31.1	35.9	42.2	36.9
DIM 2022.09		100	31.7	25.5	31.4	45.5	33.5
NI-FGSM 2022.09		100	40	35.2	39.9	56.9	43
MI-FGSM 2022.09		100	40.2	35.1	40.3	57.1	43.2
VMI-FGSM 2022.09		100	63	59.3	68.6	70.3	65.3
VNI-FGSM 2022.09		100	62.4	58.7	67.7	69.7	64.6
ANDA 2022.09		100	76.1	72.8	82.3	77	77.1
MultiANDA 2022.09		100	79.2	76	84.5	78.8	79.6
TAIG 2022.09		99.7	53.3	45.9	56.7	54.2	52.5
FIA 2022.09		98.3	78.4	75.3	81.2	83.5	79.6
MultiANDA 2022.09		93.9	87.1	85.6	99.8	84.3	87.7
ANDA 2022.09		93	86.4	83.7	99.8	82.8	86.5
FIA 2022.09		82.2	75.3	72.4	89.2	80.7	77.7
VMI-FGSM 2022.09		81.1	69.6	66.4	99.3	73.5	72.7
VNI-FGSM 2022.09		80.9	70	65.8	99.4	73.8	72.6
TAIG 2022.09		73.9	63.4	58.4	95	57.4	63.3
TIM 2022.09		64.3	35.9	30.6	25.4	70.4	45.3
NI-FGSM 2022.09		61.9	49.5	44.7	99.2	64.7	55.2
MI-FGSM 2022.09		60.3	49.3	43	98.8	64.6	54.3
SIM 2022.09		58.9	47.8	41.4	99.6	49.8	49.5
DIM 2022.09		54.6	41.4	36.6	98.2	50	45.7
TIM 2022.09		43.4	36.5	32	36	66.2	42.8
BIM 2022.09		36.3	25.6	20.6	99.3	37.8	30.1