aayush226/drone-tracking-assignment-ai-spring26

---
license: mit
task_categories:
- object-detection
language:
- en
pretty_name: UAV Drone Detection Frames
tags:
- computer-vision
- object-detection
- drone
- parquet
- image
---

# UAV Drone Detection Frames

This dataset contains video frames with at least one drone detection, stored in Parquet format.

## Contents

Each row corresponds to one saved frame and includes:

- `video`: source video name
- `frame_file`: frame filename
- `image`: embedded image bytes
- `objects.bbox`: bounding boxes in `[x, y, width, height]` format
- `objects.categories`: class ids for each detection
- `scores`: confidence scores for each detection

## Source

The dataset was created by:
1. extracting frames from input drone videos at 5 FPS,
2. running a fine-tuned YOLO detector,
3. keeping only frames with at least one detection,
4. storing the results in Hugging Face-compatible Parquet format.

## Notes

- Images are embedded directly in the Parquet file.
- Bounding boxes are stored in Hugging Face object detection format.
- This dataset was created for an academic assignment on UAV drone detection and tracking.