ReplicaOcc: A Replica-Based Benchmark for Embodied Occupancy Prediction

Replica_OCC is a benchmark dataset for evaluating embodied occupancy prediction systems, constructed in the style of EmbodiedOcc-ScanNet and OccScanNet. It provides RGB-D sequences and scene-level occupancy ground truth, released by author 'the-masses' and last updated on May 6, -2026. The ground-truth occupancy and poses are intended for evaluation-time alignment and metric computation.

Use Cases

• Benchmarking embodied occupancy prediction systems based on the provided RGB-D sequences and occupancy ground truth.
• Evaluating metric computation for 3D scene understanding based on the evaluation-time alignment data.
• Training or comparing systems for indoor scene reconstruction based on the Replica-based synthetic data organization.

Strengths

• Provides scene-level occupancy ground truth for evaluation, a feature mentioned in the description.
• Follows the data organization style of established benchmarks like EmbodiedOcc-ScanNet and OccScanNet.
• Includes RGB-D sequences, which suggests multimodal visual and depth data.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Constructed in the data organization style of EmbodiedOcc-ScanNet and OccScanNet, likely from synthetic Replica data.

Time Range

null

Freshness

Last updated 2026-05-06 03:47:55; freshness should be verified.

Geography

null