Improved Deep Point Cloud Geometry Compression: Experimental Data and Trained Models

Experimental data from the 2020 IEEE MMSP paper 'Improved Deep Point Cloud Geometry Compression' by Maurice Quach et al. The data includes all trained models, bitrates, and objective metric values (D1/D2) for multiple models and point clouds. Compressed and decompressed point clouds for models c1 to c6, G-PCC trisoup, and G-PCC octree are also provided.

Use Cases

• Benchmarking compression performance based on bitrate and D1/D2 metric data
• Comparing deep learning models (c1-c6) against G-PCC standards (trisoup, octree) using provided results
• Analyzing the relationship between bitrate and reconstruction quality using the included metric values
• Reproducing or extending the paper's experiments using the full set of trained models and point clouds

Strengths

• Includes all trained models referenced in the original 2020 IEEE publication
• Contains results for multiple models (c1 to c6, G-PCC trisoup, G-PCC octree) and multiple point clouds
• Provides both objective metric values (D1/D2) and bitrate data for performance analysis

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Last update date is unknown; freshness unverified

Provenance

Source

Maurice Quach, Laboratoire des signaux et systèmes, from the 2020 IEEE MMSP paper

Collection Method

Data generated for the experiments in the cited research paper.