ROD-Dataset: 24,326 Images for Real-Time Obstacle Detection

24,326 images annotated for 25 object classes comprise the Real-Time Obstacle Detection (ROD) dataset. It was developed at Amirkabir University of Technology, Tehran, to address sidewalk safety for pedestrians and people with visual impairments. The dataset was last updated on 2026-05-10.

Use Cases

• Train YOLO-based obstacle detection models based on the 25 annotated object classes.
• Develop smartphone-based assistive vision applications based on the dataset's focus on sidewalk safety.
• Benchmark real-time object detection performance based on the dataset's size and annotation format.
• Research urban pedestrian safety and situational awareness based on the dataset's described public-safety problem.

Strengths

• 24,326 images provide a substantial volume for model training.
• Annotations are formatted for the YOLO object detection framework.
• The dataset addresses a specific, described public-safety use case.

Limitations

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

Provenance

Source

Amirkabir University of Technology, Tehran

Collection Method

Likely collected and annotated as part of the Real-Time Obstacle Detection (ROD) project.

Freshness

Last updated 2026-05-10 18:37:27; freshness should be verified.

Geography

Tehran, Iran (inferred from the institution's location).