SSCDOL3DClassifier

This dataset was constructed to evaluate the robustness and generalization capability of 3D morphological classification methods for irregular agricultural products, using Zhacai (preserved mustard stem) as a representative example. It contains six categories of 3D Zhacai models, each captured from multiple viewing angles at 30\u00b0 rotational intervals. For every view, two corresponding 2D representations are provided: (1) a photographic image that captures appearance-based visual features and (2) a 2D scatter map that encodes the object\u2019s spatial geometry and curvature distribution. The dataset includes a total of 2,808 photographic images and 2,808 scatter representations, all normalized to a resolution of 542\u00d7539 px. This design enables researchers to analyze performance under varying perspectives and compare traditional vision-based classification with geometry-aware representations. The dataset serves as a benchmark for developing and evaluating intelligent classification methods in agricultural automation, computer vision, and 3D geometric learning.