Tag: ImageNet

Researchers introduced Cross-Distribution Semantic Transformation to enhance adversarial transferability across neural network architectures without corrupting key features or imposing excessive ...

An empirical evaluation of ten convolutional neural network architectures across three imaging modalities reveals that ImageNet pretraining gains do not ...

Researchers have shown that randomly shuffling embedded patch tokens inside vision transformer surrogate models during adversarial example generation substantially improves ...

Researchers propose SteRGAN, a GAN-based image steganography framework that achieves PSNR values above 41 dB on ImageNet and Microsoft COCO.

A new philosophical framework distinguishes remediable algorithmic bias from discrimination built into a system's representational choices, and proposes upstream ontological ...

Researchers have developed a generative adversarial network framework called StegoAdv-GAN that hides extractable secret messages inside images while simultaneously generating ...

A new study shows that popular AI explanation methods produce nearly identical heatmaps for different class labels, and introduces CASE, ...

Researchers at Tarbiat Modares University have developed DuoDiT, a dual-stream architecture that adapts large diffusion transformers for image generation by ...

Researchers have unveiled FakeMark, an attack that fabricates convincing neural network watermark evidence without ever accessing the victim model, exposing ...

Researchers have introduced NB-Net, a biologically inspired neural network that expands in width through parallel multi-scale branches and staged feature ...