DAL-PCQA dataset enables language-driven explainable point cloud quality assessment

Point Cloud Quality Assessment (PCQA) methods have traditionally predicted scalar Mean Opinion Scores (MOS) that quantify overall perceptual degradation but do not explain its causes. Human observers, by contrast, naturally reason in terms of specific distortions like blur, color shifts, point density changes, missing regions, and geometric deformations. To bridge this gap, Chakraborty et al. introduce DAL-PCQA — a distortion-aware, language-annotated dataset for PCQA, accepted at Qomex 2026. The dataset augments existing benchmark point clouds with multi-level distortion severity labels, discrete quality categories, and structured natural language descriptions that align with human perception. A key contribution is a point-cloud-specific distortion taxonomy covering both photometric and geometric artifacts, enabling fine-grained analysis of degradation patterns across distortion types and quality levels.

To validate the utility of these annotations, the authors compared zero-shot and fine-tuned multimodal models for generating perceptual quality descriptions. Results demonstrate that distortion-aware supervision substantially improves lexical and semantic alignment with ground-truth descriptions. This work enables interpretable, distortion-level reasoning, paving the way for language-driven, explainable point cloud quality assessment. The dataset is publicly available, providing a valuable resource for future research in multimodal quality assessment for 3D data.