360DVO: Real-World Omnidirectional Visual Odometry Dataset

A real-world dataset for omnidirectional visual odometry, published in IEEE Robotics and Automation Letters (RA-L) 2026. It supports the 360DVO framework, which introduces a distortion-aware spherical feature extractor and an omnidirectional differentiable bundle adjustment module. The dataset was last updated on March 25, 2026.

Use Cases

• Training deep learning models for monocular visual odometry based on omnidirectional imagery.
• Benchmarking visual odometry algorithms using distortion-aware spherical features.
• Developing and evaluating omnidirectional differentiable bundle adjustment modules.
• Researching robust pose estimation for 360-degree camera systems.

Strengths

• Dataset is associated with a paper published in a peer-reviewed journal, IEEE Robotics and Automation Letters (RA-L).
• Supports a specific deep learning framework (360DVO) with novel components like DAS-Feat and ODBA.
• Last updated on March 25, 2026, suggesting recent maintenance.

Limitations

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

Provenance

Source

chris1004336379

Collection Method

Likely collected from real-world scenes using a 360-degree camera.

Freshness

Last updated 2026-03-25 04:38:43.