CODa Re-ID: 1.04M Observations of 557 Campus Objects for Re-Identification

Approximately 1.04 million per-frame 2D observations of 557 globally unique static objects across 8 categories, captured on a university campus. The dataset was created by Dongmeyong Lee and last updated on October 15, 2025. It is intended for improving long-term robot perception and navigation under varying environmental conditions.

Use Cases

• Train object re-identification models based on globally consistent instance IDs tracked across sequences.
• Develop object-centric SLAM systems based on multi-modal annotations including 2D bounding boxes, segmentation masks, and 3D bounding boxes.
• Benchmark long-term localization algorithms based on data captured across multiple traversals under varying lighting and weather.
• Research persistent scene understanding based on observations of static objects from diverse viewpoints and distances.

Strengths

• Large scale with approximately 1.04 million annotated observations.
• Globally consistent instance IDs for 557 unique physical objects tracked across all sequences.
• Diverse capture conditions including multiple times of day, weather (sunny, overcast, rainy, low-light/night), and viewing perspectives.
• Multi-modal annotations per frame, including 2D bounding boxes, segmentation contours, projected 3D bounding boxes, and camera poses.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is limited to eight specific object categories (e.g., trees, poles, signs) from a larger original collection.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Texas Data Repository Harvested Dataverse

Collection Method

Derived by post-processing the original CODa dataset; trajectories were aligned using LiDAR-inertial odometry, objects were clustered to identify persistent 3D instances, and annotations were projected into image frames.

Time Range

null

Freshness

Last updated 2025-10-15 03:54:21; freshness should be verified.

Geography

Captured on a university campus (specific location not stated).