Weakly Supervised Domain-Specific Color Naming Based on Attention
This work addresses the need for accurate domain-specific color naming in applications like object description, but it is incremental as it builds on existing color naming methods with an attention mechanism.
The paper tackles the problem of learning domain-specific color names from weakly labeled data, which is costly to annotate, by adding an attention branch to a color naming network to modulate predictions. The method achieves state-of-the-art results on the EBAY dataset for pixel-wise and image-wise classification and can learn color names across various domains.
The majority of existing color naming methods focuses on the eleven basic color terms of the English language. However, in many applications, different sets of color names are used for the accurate description of objects. Labeling data to learn these domain-specific color names is an expensive and laborious task. Therefore, in this article we aim to learn color names from weakly labeled data. For this purpose, we add an attention branch to the color naming network. The attention branch is used to modulate the pixel-wise color naming predictions of the network. In experiments, we illustrate that the attention branch correctly identifies the relevant regions. Furthermore, we show that our method obtains state-of-the-art results for pixel-wise and image-wise classification on the EBAY dataset and is able to learn color names for various domains.