Ground-truth dataset and baseline evaluations for image base-detail separation algorithms
This work addresses a fundamental gap in computer vision research by enabling quantitative evaluation for base-detail separation, though it is incremental as it focuses on dataset creation rather than new algorithmic advances.
The authors tackled the lack of ground-truth datasets for evaluating base-detail separation algorithms in computer vision by constructing two datasets (Pascal Base-Detail with 1000 images and Fashionista Base-Detail with 250 images) and providing baseline evaluations of seven state-of-the-art methods.
Base-detail separation is a fundamental computer vision problem consisting of modeling a smooth base layer with the coarse structures, and a detail layer containing the texture-like structures. One of the challenges of estimating the base is to preserve sharp boundaries between objects or parts to avoid halo artifacts. Many methods have been proposed to address this problem, but there is no ground-truth dataset of real images for quantitative evaluation. We proposed a procedure to construct such a dataset, and provide two datasets: Pascal Base-Detail and Fashionista Base-Detail, containing 1000 and 250 images, respectively. Our assumption is that the base is piecewise smooth and we label the appearance of each piece by a polynomial model. The pieces are objects and parts of objects, obtained from human annotations. Finally, we proposed a way to evaluate methods with our base-detail ground-truth and we compared the performances of seven state-of-the-art algorithms.