Observation Sheet of Compositional Similarity Results for Impressionist Painting Pairs

This dataset contains the observation sheet with the complete set of compositional similarity results obtained during the experimental evaluation of impressionist paintings. The study was conducted on 240 manually selected pairs of visually similar impressionist paintings, derived from the ArtBench-10 dataset.For each image pair, six low-level visual transformations were independently applied: four related to color (Color Heatmap, Hue, Saturation, and Brightness) and two related to texture (Texture and Contrast). After applying each transformation, visual embeddings were extracted using a pretrained VGG-19 model, and cosine similarity was computed between the corresponding embeddings of each pair.The observation sheet reports the normalized similarity scores (range [0,1]) for each transformation and each image pair. These results constitute the basis for the descriptive and inferential statistical analyses presented in the associated research paper, including one-way ANOVA and Tukey post-hoc tests.This resource is provided to support reproducibility and further research in computational art analysis, visual similarity assessment, and content-based image retrieval.