DepthVLM-Bench: A Unified Indoor-Outdoor Metric Depth Benchmark

DepthVLM-Bench is a unified metric depth estimation benchmark for vision-language models. It provides diverse indoor and outdoor scenes with metric depth annotations in a VLM-compatible format. The dataset was created by JonnyYu828 and was last updated on May 18, 2026.

Use Cases

• Training vision-language models for metric depth prediction based on unified indoor-outdoor scenes.
• Benchmarking model performance on dense geometry tasks using the provided depth supervision.
• Developing multimodal systems that combine depth estimation with language understanding based on the VLM-compatible format.

Strengths

• Specifically designed for vision-language models with a unified, VLM-compatible data format.
• Provides metric depth annotations for both indoor and outdoor scenes, offering diverse geometric contexts.
• Last updated on May 18, 2026, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Freshness

Last updated 2026-05-18 03:58:02