OliseNS/person-face-package-home-security-detection

---
license: other
license_name: multi-source-aggregated-terms
license_link: LICENSE
pretty_name: "Person, Face & Package — Home Security Detection (YOLO)"
task_categories:
- object-detection
tags:
- image
- object-detection
- computer-vision
- yolo
- ultralytics
- bbox
- home-security
- surveillance
- face-detection
- parcel-detection
- delivery
- vehicle-detection
- pet-detection
- bird-detection
annotations_creators:
- crowdsourced
- found
language:
- en
---

# Person, face & package — home-security detection dataset

YOLO-format dataset for person, face, vehicles, small vehicles, parcels, pets, and birds in home / delivery / street scenes. Full data lives under `bigsplit/`; `sample/` is a 100-image preview subset (same layout: flat `images/` and `labels/`).

Training configs point at a `data.yaml` beside those folders. Both splits use the same class list (`nc` and `names`); only the root path and which image set is packaged differ. Labels are YOLO `.txt` files with one line per box: `class_id x_center y_center width height` (normalized 0–1).

## `data.yaml` — tags and layout

Class **tags** are the `names` list; order is the YOLO class index (`0` … `nc - 1`). Below is the canonical schema used by `sample/data.yaml` (checked in). The full split uses the same `nc` and `names` in `bigsplit/data.yaml`; `path`, `train`, and `val` resolve to that folder’s `images/` when you train from the dataset root.

**`sample/data.yaml`**

```yaml
# Preview sample (100 images); same layout as bigsplit — no split subfolders.
path: .
train: images
val: images
nc: 9
names:
- person
- face
- vehicle
- bike
- bicycle
- parcel
- dog
- cat
- bird
```

When the full dataset is present, `bigsplit/data.yaml` uses the same `path` / `train` / `val` / `nc` / `names` (often with a file header comment describing the full split instead of the sample).

## Preview (bounding boxes)

Five PNGs are generated from the checked-in **`sample/`** images and YOLO labels. Each file stitches **three** frames **edge-to-edge** at equal height (**no padding** between panels). Boxes and class names are drawn from the corresponding `sample/labels/*.txt` files.

<p align="center">
<img src="assets/preview_01.png" width="32%" alt="Sample preview 1 — three stitched frames" />
<img src="assets/preview_02.png" width="32%" alt="Sample preview 2 — three stitched frames" />
<img src="assets/preview_03.png" width="32%" alt="Sample preview 3 — three stitched frames" />
</p>
<p align="center">
<img src="assets/preview_04.png" width="32%" alt="Sample preview 4 — three stitched frames" />
<img src="assets/preview_05.png" width="32%" alt="Sample preview 5 — three stitched frames" />
</p>

[preview_01.png](assets/preview_01.png) · [preview_02.png](assets/preview_02.png) · [preview_03.png](assets/preview_03.png) · [preview_04.png](assets/preview_04.png) · [preview_05.png](assets/preview_05.png) · [preview_manifest.txt](assets/preview_manifest.txt) (source filenames per composite)

## Layout

- `bigsplit/data.yaml` — class list and paths (`train` / `val` both point at `images/`).
- `sample/data.yaml` — same schema for the preview subset.

## Classes (`nc: 9`)

| id | name |
|---:|----------|
| 0 | person |
| 1 | face |
| 2 | vehicle |
| 3 | bike |
| 4 | bicycle |
| 5 | parcel |
| 6 | dog |
| 7 | cat |
| 8 | bird |

## License

See `LICENSE`.

---
许可证：其他
许可证名称：多源聚合条款
许可证链接：LICENSE
数据集展示名："人物、面部与包裹——家庭安全检测（YOLO）"
任务类别：
- 目标检测（object detection）
标签：
- 图像
- 目标检测（object detection）
- 计算机视觉（computer vision）
- YOLO
- Ultralytics
- 边界框（bbox）
- 家庭安全
- 监控
- 人脸检测（face detection）
- 包裹检测（parcel detection）
- 配送
- 车辆检测（vehicle detection）
- 宠物检测（pet detection）
- 鸟类检测（bird detection）
标注创作者：
- 众包（crowdsourced）
- 公开获取
语言：
- 英语
---

# 人物、面部与包裹——家庭安全检测数据集

这是一个适配YOLO格式的数据集，涵盖家庭、配送、街道场景中的人物、面部、机动车、非机动车、包裹、宠物与鸟类。完整数据集存放于`bigsplit/`目录下；`sample/`为包含100张图像的预览子集（目录结构一致：均为扁平化的`images/`与`labels/`文件夹）。

训练配置需指向对应文件夹旁的`data.yaml`文件。两个数据集子集使用完全相同的类别列表（即`nc`与`names`字段），仅根目录路径与打包的图像集存在差异。标注文件为YOLO格式的`.txt`文本，每个边界框对应一行，格式为：`class_id x_center y_center width height`（坐标归一化至0~1区间）。

## `data.yaml`：标签与目录结构

类别**标签**即`names`字段对应的列表，其顺序即为YOLO的类别索引（从`0`到`nc - 1`）。以下为`sample/data.yaml`的标准配置模板（已提交至代码仓库）。完整数据集的`bigsplit/data.yaml`使用完全相同的`nc`与`names`字段；若从数据集根目录启动训练，`path`、`train`与`val`字段将指向该目录下的`images/`文件夹。

**`sample/data.yaml`**

yaml
# Preview sample (100 images); same layout as bigsplit — no split subfolders.
path: .
train: images
val: images
nc: 9
names:
- person
- face
- vehicle
- bike
- bicycle
- parcel
- dog
- cat
- bird

当使用完整数据集时，`bigsplit/data.yaml`的`path`、`train`、`val`、`nc`与`names`字段配置完全一致，仅文件头部注释会说明完整数据集而非预览子集。

## 预览效果（边界框可视化）

基于已提交至仓库的**`sample/`**目录下的图像与YOLO标注，共生成5张PNG格式预览图。每张图片将**3帧图像按等高度无缝拼接**（面板间无留白），边界框与类别名称均从对应`sample/labels/*.txt`标注文件中读取。

<p align="center">
<img src="assets/preview_01.png" width="32%" alt="样本预览1 — 三张拼接帧" />
<img src="assets/preview_02.png" width="32%" alt="样本预览2 — 三张拼接帧" />
<img src="assets/preview_03.png" width="32%" alt="样本预览3 — 三张拼接帧" />
</p>
<p align="center">
<img src="assets/preview_04.png" width="32%" alt="样本预览4 — 三张拼接帧" />
<img src="assets/preview_05.png" width="32%" alt="样本预览5 — 三张拼接帧" />
</p>

[preview_01.png](assets/preview_01.png) · [preview_02.png](assets/preview_02.png) · [preview_03.png](assets/preview_03.png) · [preview_04.png](assets/preview_04.png) · [preview_05.png](assets/preview_05.png) · [preview_manifest.txt](assets/preview_manifest.txt)（每张合成图对应的源文件名列表）

## 目录结构

- `bigsplit/data.yaml`：包含类别列表与路径配置（`train`与`val`均指向`images/`文件夹）。
- `sample/data.yaml`：预览子集使用相同的配置模板。

## 类别列表（`nc: 9`）

| 类别ID | 类别名称 |
|-------:|------------|
| 0 | 人物 |
| 1 | 面部 |
| 2 | 机动车 |
| 3 | 小型自行车 |
| 4 | 普通自行车 |
| 5 | 包裹 |
| 6 | 狗 |
| 7 | 猫 |
| 8 | 鸟类 |

## 许可证

详见`LICENSE`文件。