Flexible ViG: Learning the Self-Saliency for Flexible Object Recognition | Read Paper on Bytez