Benchmark RadarRDVSv2
RDVSv2 is a large-scale benchmark for RGB-D video salient object detection, containing 249 video sequences with 29,077 annotated frames. It includes depth maps, optical flow, and eye-tracking-guided salient object masks. The benchmark provides a fixed dataset and evaluation protocol for comparing models on this task.
- Released
- 2026-07-28
- Readiness
- Runnable
- Primary field
- General AI
Why it matters
Existing RGB-D VSOD datasets are limited in scale and annotation quality, hindering progress. RDVSv2 offers a larger, more diverse, and challenging benchmark, enabling more robust evaluation and comparison of models, and supporting the development of methods that can handle real-world scenarios.
Motivation
We introduce RDVSv2, a large-scale benchmark for RGB-D video salient object detection (RGB-D VSOD) with dense frame-level annotations.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.