Are visual dictionaries generalizable? by Otavio A. B. Penatti, Eduardo Valle, and Ricardo da S. Torres
Mid-level features based on visual dictionaries are today a cornerstone of systems for classification and retrieval of images. Those state-of-the-art representations depend crucially on the choice of a codebook (visual dictionary), which is usually derived from the dataset. In general-purpose, dynamic image collections (e.g., the Web), one cannot have the entire collection in order to extract a representative dictionary. However, based on the hypothesis that the dictionary reflects only the diversity of low-level appearances and does not capture semantics, we argue that a dictionary based on a small subset of the data, or even on an entirely different dataset, is able to produce a good representation, provided that the chosen images span a diverse enough portion of the low-level feature space. Our experiments confirm that hypothesis, opening the opportunity to greatly alleviate the burden in generating the codebook, and confirming the feasibility of employing visual dictionaries in large-scale dynamic environments.
The authors use the Caltech-101 image set because of its “diversity.” Odd because they cite the Caltech-256 image set, which was created to answer concerns about the lack of diversity in the Caltech-101 image set.
Not sure this paper answers the issues it raises about visual dictionaries.
Wanted to bring it to your attention because representative dictionaries (as opposed to comprehensive ones) may be lurking just beyond the semantic horizon.