MUFASA：面向槽注意力的多层框架
MUFASA: A Multi-Layer Framework for Slot Attention
把槽注意力从只在ViT最后一层做，扩展到多层特征上分别做再融合，这个即插即用模块可以直接套到现有无监督目标分割/目标中心表征方法上，用极小的推理开销换取分割提升和更快收敛。
原文依据 we introduce MUFASA, a lightweight plug-and-play framework for slot-attention-based approaches to unsupervised object segmentation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论无监督目标中心学习（OCL）将视觉场景分解为不同实体。槽注意力是一种流行方法，它将单个物体表示为称为槽的潜变量向量。现有方法仅从预训练视觉Transformer（ViT）的最后一层获得这些槽表示，忽略了其他层中编码的有价值且语义丰富的信息。为更好地利用这些潜变量语义信息，我们提出MUFASA，一个用于基于槽注意力的无监督目标分割方法的轻量级即插即用框架。我们的模型在ViT编码器的多个特征层上计算槽注意力，充分利用其语义丰富性。我们提出一种融合策略，将多层上获得的槽聚合为统一的目标中心表示。将MUFASA集成到现有OCL方法中，可在多个数据集上改善其分割结果，在仅带来少量推理开销的同时提升训练收敛，并达到新的最先进水平。
英文原文摘要 1057 字符
Unsupervised object-centric learning (OCL) decomposes visual scenes into distinct entities. Slot attention is a popular approach that represents individual objects as latent vectors, called slots. Current methods obtain these slot representations solely from the last layer of a pre-trained vision transformer (ViT), ignoring valuable, semantically rich information encoded across the other layers. To better utilize this latent semantic information, we introduce MUFASA, a lightweight plug-and-play framework for slot-attention-based approaches to unsupervised object segmentation. Our model computes slot attention across multiple feature layers of the ViT encoder, fully leveraging their semantic richness. We propose a fusion strategy to aggregate slots obtained on multiple layers into a unified object-centric representation. Integrating MUFASA into existing OCL methods improves their segmentation results across multiple datasets, setting a new state of the art while simultaneously improving training convergence with only minor inference overhead.