DMformer: a transformer with denoising and multi-modal data fusion for enhancing BEV perception

DMFormer is introduced, an innovative multi-modal BEV perception framework that employs Transformer architecture and a diffusion denoising model to tackle key challenges, including sensor noise, efficient fusion of multi-modal data, and modeling dynamic scenes.