Hugging Face's logo

Datasets:
LK-ROBOTICS
/
nxp-sf-train-balanced

Tasks:
Object Detection
Modalities:
Image
Languages:
Korean
Tags:
image
yolo
patrol
safety
fire-detection
smoke-detection
License:
Dataset card Files
xet
 Community
nxp-sf-train-balanced
File size: 2,632 Bytes
3e85e89
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
---
task_categories:
  - object-detection
license: other
tags:
  - image
  - yolo
  - patrol
  - safety
  - fire-detection
  - smoke-detection
  - sf
pretty_name: "NXP SF Balanced Fire/Smoke Dataset v2"
language:
  - ko
size_categories:
  - 10K<n<100K
---

# NXP SF Balanced Fire/Smoke Dataset v2

## 개요

`nxp_sf_train_balanced_v2`는 NXP `lk_vision_patroller` 전면 화재/연기 탐지 후보 모델을 학습하기 위한 YOLO 형식 2-class SF(smoke/fire) 데이터셋 revision v2입니다.

이 revision은 `balanced-v1` backbone을 유지하면서, 최신 버전 우선 규칙에 따라 D-family는 `roboflow_psf_train_d_v3`만 반영하고, `e_lighter` fire bias를 줄인 구성입니다.

클래스 스키마는 다음으로 고정합니다.

```yaml
0: fire
1: smoke
```

## 구성

- 포맷: YOLO detection (`images/` + `labels/*.txt`)
- split:
  - `train/images`, `train/labels`
  - `val/images`, `val/labels`
  - `test/images`, `test/labels`
- 총 이미지 수: 11,936
- 총 box 수:
  - fire: 11,800
  - smoke: 8,785
- empty-label 이미지 수: 2,511

| split | images | fire boxes | smoke boxes | empty-label images |
|---|---:|---:|---:|---:|
| train | 9,277 | 9,156 | 6,750 | 2,003 |
| val | 1,493 | 1,541 | 1,175 | 260 |
| test | 1,166 | 1,103 | 860 | 248 |

## 주요 변화 (v1 -> v2)

- 유지: `kaggle_sf_train_dfire_sampled5k_v1`, `roboflow_sf_train_a_v1`, `roboflow_sf_train_b_v1`
- 유지: `roboflow_sf_train_anguk_v1`, `roboflow_sf_train_daegu_v1`
- 조정: `roboflow_sf_train_gongduk_v1` 비중 소폭 증가
- 추가: `roboflow_psf_train_d_v3`
- 추가: `roboflow_psf_train_gongduk_b_v2`
- 추가: `roboflow_psf_train_e_lighter_v1` (축소 반영)
- 최신 버전 우선: D-family는 `v1`, `v2`를 제외하고 `v3`만 사용

## Hugging Face revision policy

This dataset family uses a stable repo id plus revision paths.

- stable repo: `LK-ROBOTICS/nxp-sf-train-balanced`
- this revision path: `revisions/v2`

## 사용법

Ultralytics YOLO 학습 예시:

```python
from ultralytics import YOLO

model = YOLO("yolo11n.pt")
model.train(
    data="/home/jinhyuk2me/lk_ws/ai/vision/perception/projects/patroller-sf-yolo11n-nxp/data-nxp-sf-train-balanced-v2.yaml",
    imgsz=320,
    epochs=200,
    patience=40,
    batch=192,
    workers=8,
)
```

## Provenance

- dataset id: `nxp_sf_train_balanced_v2`
- build manifest: `BUILD_MANIFEST.yaml`
- MMS recipe metadata: `/home/jinhyuk2me/lk_ws/ops/ml-management/lk_mms/datasets/recipes/nxp-sf-train-balanced-v2.yaml`
- project YAML: `/home/jinhyuk2me/lk_ws/ai/vision/perception/projects/patroller-sf-yolo11n-nxp/data-nxp-sf-train-balanced-v2.yaml`