Object-Centric Learning with Slot Attention
Francesco Locatello Dept. of Computer Science, ETH Zurich Max-Planck Institute for Intelligent Systems Dirk Weissenborn Google Research, Brain Team Thomas Unterthiner Google Research, Brain Team Aravindh Mahendran Google Research, Brain Team Georg Heigold Google Research, Brain Team Jakob Uszkoreit Google Research, Brain Team Alexey Dosovitskiy Google Research, Brain Team Thomas Kipf Google Research, Brain Team
Abstract
Learning object-centric representations of complex scenes is a promising step towards enabling efficient abstract reasoning from low-level perceptual features. Yet, most deep learning approaches learn distributed representations that do not capture the compositional properties of natural scenes. In this paper, we present the Slot Attention module, an architectural component that interfaces with perceptual representations such as the output of a convolutional neural network and produces a set of task-dependent abstract representations which we call slots. These slots are exchangeable and can bind to any object in the input by specializing through a competitive procedure over multiple rounds of attention. We empirically demonstrate that Slot Attention can extract object-centric representations that enable generalization to unseen compositions when trained on unsupervised object discovery and supervised property prediction tasks.
中文速览
如何让神经网络像人类一样把一张图像自动拆解成一个个独立的"物体"，是视觉理解领域的核心难题。这篇论文提出了"槽位注意力"（Slot Attention）模块：它接收卷积网络提取的图像特征，通过多轮竞争式注意力迭代，将这些特征动态分配给若干个结构相同、可互换的"槽位"（slot），每个槽位自动绑定到图像中的某个物体，无需预先指定每个槽位负责哪类物体。实验表明，在无监督物体发现任务中该方法达到或超越了现有最优方法，同时训练速度更快、内存占用更少；在有监督的物体属性预测任务中，模型无需分割标注就能自动聚焦到各个物体上。这一模块的意义在于它提供了一个通用、轻量、易于即插即用的组件，使模型能够学到真正以物体为中心的表示，并自然地泛化到训练时未见过的物体组合，为后续视觉推理、物理仿真等需要结构化理解的任务奠定了基础。
原文 arXiv:2006.15055;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2006.15055v2