論文閱讀之 Inferring Analogous Attributes CVPR 2014

Inferring Analogous Attributes     CVPR  2014

Chao-Yeh Chen and Kristen Grauman

Abstract:

The appearance of an attribute can vary considerably from class to class (e.g., a “fluffy” dog vs. a “fluffy” towel), making standard class-independent attribute models break down. Yet, training object-specific models for each attribute can be impractical, and defeats the purpose of using attributes to bridge category boundaries. We propose a novel form of transfer learning that addresses this dilemma. We develop a tensor factorization approach which, given a sparse set of class-specific attribute classifiers, can infer new ones for object-attribute pairs unobserved during training. For example, even though the system has no labeled images of striped dogs, it can use its knowledge of other attributes and objects to tailor “stripedness” to the dog category. With two large-scale datasets, we demonstrate both the need for category-sensitive attributes as well as our method’s successful transfer. Our inferred attribute classifiers perform similarly well to those trained with the luxury of labeled class-specific instances, and much better than those restricted to traditional modes of transfer.

從上圖可以看出，通過學習一些特定目标的屬性分類器，我們可以類推出相似的屬性分類器．該分類器是對目标敏感的，雖然沒有特定種類的帶标簽的訓練圖像．

１．Introduction:

本文的核心貢獻有３點：

１．First, performing transfer jointly in the space of two labeled aspects of the data—namely, categories and attributes—is new. Critically, this means our method is not confined to transfer along same-object or same-attribute boundaries; rather, it discovers analogical relationships based on some mixture of previously seen objects and attributes.

第一點，就是與傳統的轉移學習不同，本文的轉移是聯合的轉移，即：目标種類和屬性的轉移．

２．Second, our approach produces a discriminative model for an attribute with zero training examples from that category.

第二點，就是産生一種判别性的模型，盡管該類屬性沒有訓練樣本．

３．Third, while prior methods often require information about which classes should transfer to which [2, 29, 26, 1] (e.g., that a motorcycle detector might transfer well to a bicycle), our approach naturally discovers where transfer is possible based on how the observed attribute models relate. It can transfer easily between multiple classes at once, not only pairs, and we avoid the guesswork of manually specifying where transfer is likely.

第三點，就是本文所提出的方法不需要關于什麼轉移到什麼的資訊．而可以在多種類别之間很友善的轉移．

2. Related Work

In contrast, our approach implicitly discovers analogical relationships among object-sensitive attribute classifiers, and our goal is to generate

novel category-sensitive attribute classifiers.

3. Approach

Given training images labeled by their category and one or more attributes, our method produces as output a series of category-sensitive attribute classifiers. Some of those classifiers are explicitly trained with the labeled data, while the rest are inferred by our method. We show how to create these analogous attribute classifiers via tensor completion.

In the following, we first describe how we train category-sensitive classifiers (Sec. 3.1). Then we define the tensor of attributes (Sec. 3.2) and show how we use it to infer analogous models (Sec. 3.3). Finally, we discuss certain salient aspects of the method design (Sec. 3.4).

3.1. Learning Category-Sensitive Attributes

在現有的系統當中，屬性的訓練是通過一種種類之間互相獨立的方式 ( in a category-independent manner )進行．

在這個工作中，我們挑戰傳統的訓練方式，即：in a completely category-indenpent mannner.

while attributes’ visual cues are often shared among some objects, the sharing is not universal. It can dilute(稀釋) the learning process to pool cross-category exemplars indiscriminately.　（在某些物體中，屬性的視覺線索通常是共享的，但是這種共享不是普遍的．能夠非判别性的稀釋學習過程來集中跨種類的樣本）．

一種比較 naive 的做法就是，instead train category-sensitive attributes would be to partition training exemplars by their category labels, and train one attribute per category.　當有足夠的 attribute + object combinations 的帶标簽的樣本時，這種政策可能是足夠的．但是，初步實驗證明該方法是次于訓練單個普遍的屬性．我們歸結了兩點原因：

１．even in large-scale collections, the long-tailed distribution of object/scene/attribute occurrences in the real world means that some label pairs will be undersampled, leaving inadequate exemplars to build a statistically sound model,

2．this naive approach completely ignores attributes’ inter-class semantic ties.　屬性類别之間的語意連接配接．

To overcome these shortcomings, we instead use an importance-weighted support vector machine (SVM) to train each category-sensitive attribute. 每一個訓練樣本(xi, yi)都包括一個圖像描述xi，和标簽yi 屬于{-1, +1}.