MixSA: Training-free Reference-based Sketch Extraction via Mixture-of-Self-Attention

DiffMRe-assign community