PotriAbhiB/uav-drone-detection-sample-frames

---
pretty_name: UAV Drone Detection Sample Frames
task_categories:
- object-detection
language:
- en
tags:
- computer-vision
- drones
- object-detection
---

# UAV Drone Detection Sample Frames

This dataset contains sample annotated detection frames produced for a UAV drone detection and tracking assignment.

## Files

- `drone_detections.parquet`: one row per detected frame sample.

## Parquet schema

- `image`: annotated detection frame
- `video_name` (string): source video folder name
- `frame` (string): frame filename
- `x1` (float): left bounding box coordinate
- `y1` (float): top bounding box coordinate
- `x2` (float): right bounding box coordinate
- `y2` (float): bottom bounding box coordinate
- `center_x` (float): detected bounding box center x-coordinate
- `center_y` (float): detected bounding box center y-coordinate
- `confidence` (float): detector confidence score

## Notes

This is a sample output dataset generated from the assignment pipeline, not the original training dataset.

The detections were produced using a fine-tuned YOLOv8n drone detector. Bounding boxes, center coordinates, and confidence scores were extracted from the detector output and saved into Parquet format for submission.

## Source videos

- `drone_video_1`
- `drone_video_2`

## Sampling

- This dataset contains a sample of detection frames from the full assignment output.
- Up to 100 detection samples were included per video for a compact Parquet deliverable.

## Quick usage

```python
import pandas as pd

df = pd.read_parquet("drone_detections.parquet")
print(df.head())