OpenScene-Classification
收藏魔搭社区2025-12-03 更新2025-05-24 收录
下载链接:
https://modelscope.cn/datasets/prithivMLmods/OpenScene-Classification
下载链接
链接失效反馈官方服务:
资源简介:
# OpenScene-Classification Dataset
A high-quality image classification dataset curated for **scene detection tasks**, particularly useful in training and evaluating models for recognizing various natural and man-made environments.
## Dataset Summary
The **OpenScene-Classification** dataset contains labeled images categorized into six distinct scene types:
- `buildings`
- `forest`
- `glacier`
- `mountain`
- `sea`
- `street`
This dataset is structured for supervised image classification, suitable for deep learning models aiming to identify and classify real-world scenes.
## Dataset Structure
- **Split:** `train` (currently only one split)
- **Format:** `parquet`
- **Modality:** `Image`
- **Labels Type:** Integer class indices with corresponding string names
- **License:** Apache-2.0
Each entry in the dataset includes:
- `image`: the image of the scene
- `label`: the class index (e.g., 0 for buildings)
- `label_name`: the class name (e.g., "buildings")
> Note: The dataset viewer on Hugging Face may take a moment to load all samples.
## Label Mapping
| Class Index | Label |
|-------------|------------|
| 0 | buildings |
| 1 | forest |
| 2 | glacier |
| 3 | mountain |
| 4 | sea |
| 5 | street |
## Dataset Stats
- **Size**: 10K - 100K images
- **Language**: English (tags, metadata)
- **Tags**: `Scene-Detection`, `buildings`, `forest`, `glacier`, `mountain`, `sea`, `street`
## Intended Use
This dataset is ideal for:
- Scene classification model training
- Benchmarking computer vision algorithms
- Educational purposes in machine learning and AI
# OpenScene-Classification 数据集(OpenScene-Classification Dataset)
这是一款专为**场景检测任务(Scene-Detection Tasks)**打造的精选高质量图像分类数据集,可广泛用于训练与评估各类用于识别自然与人工营造环境的模型。
## 数据集概览
**OpenScene-Classification 数据集**包含带标注的图像,共分为6种不同的场景类别:
- 建筑物(buildings)
- 森林(forest)
- 冰川(glacier)
- 山地(mountain)
- 海洋(sea)
- 街道(street)
本数据集专为监督式图像分类任务设计,适配各类旨在识别并分类真实世界场景的深度学习模型。
## 数据集结构
- **数据划分**:仅包含训练集(train)一个划分
- **存储格式**:Parquet(parquet)
- **模态**:图像(Image)
- **标签类型**:带对应字符串名称的整数类别索引
- **授权协议**:Apache-2.0
数据集的每条数据包含以下字段:
- `image`:场景图像
- `label`:类别索引(例如建筑物对应索引0)
- `label_name`:类别名称(例如"buildings")
> 注意:Hugging Face平台上的数据集查看器可能需要些许时间才能加载全部样本。
## 标签映射
| 类别索引 | 标签名称(英文原文) |
|---------|----------------------|
| 0 | buildings(建筑物) |
| 1 | forest(森林) |
| 2 | glacier(冰川) |
| 3 | mountain(山地) |
| 4 | sea(海洋) |
| 5 | street(街道) |
## 数据集统计信息
- **规模**:1万至10万张图像
- **语言**:英语(用于标签、元数据)
- **标签**:场景检测(Scene-Detection)、建筑物(buildings)、森林(forest)、冰川(glacier)、山地(mountain)、海洋(sea)、街道(street)
## 适用场景
本数据集适用于:
- 场景分类模型的训练
- 计算机视觉算法的基准测试
- 机器学习与人工智能领域的教学实践
提供机构:
maas
创建时间:
2025-05-22



