Automatic discovery of discriminative parts as a quadratic assignment problem
Sicre, Rabin, Avrithis, Furon, Jurie · cs.CV · 2016-11-14 · 原文
Part-based image classification consists in representing categories by small sets of discriminative parts upon which a representation of the images is built. This paper addresses the question of how to automatically learn such parts from a set of labeled training images. The training of parts is cast as a quadratic assignment problem in which optimal correspondences between image regions and parts are automatically learned. The paper analyses different assignment strategies and thoroughly evaluates them on two public datasets: Willow actions and MIT 67 scenes. State-of-the art results are obtained on these datasets.
讲义
讲义·推断 依据「原文」自动生成的结构化摘要（推断），非原文表述；以原文为准。
1. 人话版
Part-based image classification consists in representing categories by small sets of discriminative parts upon which a representation of the images is built.
This paper addresses the question of how to automatically learn such parts from a set of labeled training images.
2. 领域脉络
本文类目：cs.CV，属于其所在研究脉络的最新进展。
3. 机制拆解
The training of parts is cast as a quadratic assignment problem in which optimal correspondences between image regions and parts are automatically learned.
State-of-the art results are obtained on these datasets.
4. 证据与数字
The paper analyses different assignment strategies and thoroughly evaluates them on two public datasets: Willow actions and MIT 67 scenes.
5. 反例与边界
摘要未声明局限与反例——这是需要警惕的信号，精读时先问边界。
6. 跨领域连接与意外收获
思考本文机制能否迁移到你正在跟进的问题。
7. 可复用方法
把本文机制与你手头项目对照，找一个两周内能验证的最小实验。
8. 术语表
精读时把不熟的术语记入此处，作为下次回忆的锚点。