Samer P. Francy, MSc, CSE, PMP’s Post

A compressed ConvNeXt model and CIFAR-10 dataset were used to test the effect of adversarial attacks using FGSM (Fast Gradient Sign Method). The accuracy dropped from 93% to 24%!!! and the below sample shows how the attack mislead the model to classify frog as a cat!! #artificialintelligence #adversarialattacks #convnext #fgsm

  • chart
Mohammed AlHaj

Driving Automation & Digital Transformation at etisalat by e& | AI Champion & Researcher | MSc/AI, University of Bath

2mo

No enough attention to this!!

Like
Reply

To view or add a comment, sign in

Explore topics