rollingcoconut/drone-detection-assignment3
收藏Hugging Face2026-03-23 更新2026-03-29 收录
下载链接:
https://hf-mirror.com/datasets/rollingcoconut/drone-detection-assignment3
下载链接
链接失效反馈官方服务:
资源简介:
---
title: Drone Detection and Tracking Assgn 3
dataset_info:
features:
- name: video
dtype: string
- name: frame_index
dtype: int64
- name: image
dtype: image
- name: num_detections
dtype: int64
- name: detections
dtype: string
splits:
- name: train
num_bytes: 146107544.0
num_examples: 729
download_size: 141974498
dataset_size: 146107544.0
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
---
# UAV Drone Detection and Tracking
Detect and track drones in video using YOLO fine-tuned detector, Kalman Filter w/ tuned noise parameters.
## Source Videos
| Video | URL |
|-------|-----|
| drone_video_1 | [YouTube - DhmZ6W1UAv4](https://www.youtube.com/watch?v=DhmZ6W1UAv4) |
| drone_video_2 | [YouTube - YrydHPwRelI](https://www.youtube.com/watch?v=YrydHPwRelI) |
## Training Dataset
* **Dataset:** [lgrzybowski/seraphim-drone-detection-dataset](https://huggingface.co/datasets/lgrzybowski/seraphim-drone-detection-dataset)
* **Training images used:** 33296 (from ~77K )
* **Test images:** 8349
## Detector
* **Model:** YOLOv8n fine-tuned on Seraphim Drone Detection Dataset
* **Training:** 15 epochs, 640x640, batch 16
* **Classes:** 1 (drone)
* **Performance:** Precision 0.93, Recall 0.863
* **Frame extraction:** 5 fps via ffmpeg
## Schema
### `detections` — Frame-Level Drone Detections
One row per video frame that contains at least one drone detection. Confidence threshold >= 0.5.
| Column | Type | Description |
|--------|------|-------------|
| video | string | Video name (drone_video_1, drone_video_2) |
| frame_index | int | Frame number within the video (0-indexed at 5 fps) |
| image | bytes | The video frame image (JPEG) |
| num_detections | int | Number of drones detected in this frame |
| detections | string (JSON list) | List of detections with bbox and confidence |
Example `detections` value:
```json
[{"bbox": [839.67, 173.83, 919.57, 229.07], "confidence": 0.8177}]
```
NOTE: `[x_min, y_min, x_max, y_max]` pixel coordinates.
## Detection Summary
| Video | Total Frames | Frames with Detections | Detection Rate |
|-------|-------------|----------------------|----------------|
| drone_video_1 | 828 | 551 | 66.5% |
| drone_video_2 | 2,580 | 178 | 6.9% |
提供机构:
rollingcoconut



