AI-TOD: Tiny Object Detection in Aerial Images

AI-TOD provides aerial imagery for tiny object detection, released by jwwangchn and updated in November 2024. It focuses on identifying objects with extremely small pixel footprints in high-altitude captures.

Use Cases

• Training detection models for objects with minimal pixel area in aerial frames
• Benchmarking localization accuracy for dense, small-scale object clusters
• Evaluating the performance of scale-invariant feature extraction layers

Strengths

• MIT license for open redistribution and modification
• Specialized focus on objects with very small pixel dimensions
• Maintained repository with updates as of late 2024

Limitations

• Total record count and specific class labels are not specified in the provided metadata
• High potential for label noise given the difficulty of annotating sub-pixel objects

Provenance

Source

jwwangchn via GitHub

Collection Method

annotated

Freshness

Last updated November 2024.