Nours999/intro-ai-assignment-3

# intro-ai-assignment-3: UAV Drone Detection Dataset

## Overview
Detection results from YOLO26 model deployed on UAV tracking videos.

**Model trained on:** UAV Detection Dataset (Roboflow Universe)
**Dataset reference:** https://universe.roboflow.com/uav-i6z6x/uav-detection-zlcin
**Total detections:** 530 across 495 frames

## Schema (detections.parquet)

| Column | Type | Description |
|--------|------|-------------|
| `video_id` | str | Name of video source |
| `frame_index` | int | Frame sequence number (0-indexed) |
| `timestamp_sec` | float | Video timestamp in seconds (extracted at 5 fps) |
| `class_label` | str | Detected object category (e.g., 'drone') |
| `confidence_score` | float | YOLO prediction confidence [0.0, 1.0] |
| `bounding_box` | str | Format: (x_min, y_min, x_max, y_max) in pixels |

## Usage Example

```python
import pandas as pd

df = pd.read_parquet("detections.parquet")

# Find all high-confidence drone detections
drones = df[(df['class_label'] == 'drone') & (df['confidence_score'] > 0.6)]
print(f"{len(drones):,} high confidence detections")

# Get detections at specific timestamp
t60 = df[df['timestamp_sec'] == 60.0]
print(f"At t=60s: {len(t60)} detections")
```

---
Generated from assignment-3 YOLO drone tracking pipeline.