List your company

Audio-visual event recognition on Kinetics-Sounds (test)

72.71AV ScoreExFMem
34.521244.435654.3564.2644Apr 5, 2021

Evaluation results

MethodLinks
ExFMemDefense=External feature memory banks (proposed)2021.0472.7141.5629.9310.4427.313.99
NoneDefense=No defense baseline2021.0472.4236.426.358.0923.610
NoneAttacker=MIM, Attack Strength=0.062021.0472.4211.189.63327.040
MaxSimDefense=Maximizing audio-visual similarity2021.0471.3934.9529.5721.4628.664.02
MaxSimAttacker=MIM, Attack Strength=0.062021.0471.3917.6518.781,38916.778.7
MinSim+ExFMemDefense=Full defense model (proposed)2021.0471.3355.9630.5724.937.1412.44
MinSim + ExFMem (Full)Attacker=MIM, Attack Strength=0.062021.0471.3344.7216.0499923.5815.45
MinSimDefense=Dissimilarity constraint (proposed)2021.0470.8852.4228.1221.6234.058.99
Unimodal VDefense=Visual-only network2021.0466.0866.0818.7218.7234.54.55
Unimodal VAttacker=MIM, Attack Strength=0.062021.0466.0866.085.7757725.8712.49
PCLDefense=Prototype conformity loss2021.0464.563.4329.2828.6740.468.93
PCLAttacker=MIM, Attack Strength=0.062021.0464.562.9820.261,99734.419.44
Unimodal ADefense=Audio-only network2021.0435.991.8735.991.8713.24-46.8
Unimodal AAttacker=MIM, Attack Strength=0.062021.0435.990.135.991012.06-31.41