The Little Red Book multimodal team proposes a new method a Controlled Brain Visual Diffusion Model (CMVDM). The model can well reduce fMRI signals into high-quality pictures with semantic properties consistent with the original visual stimulus and spatial structure alignment, which can make the generated images more clear and closer to the original images seen by the human eye.
Little Red Book's Multimodal Team Creates New "Diffusion Model"
Previous: AMD首批Instinct MI300X已开始交付
Next: 全国首个国产液冷万卡算力集群启动建设