five

Multilabel-Portrait-18K

收藏
魔搭社区2025-12-03 更新2025-04-26 收录
下载链接:
https://modelscope.cn/datasets/prithivMLmods/Multilabel-Portrait-18K
下载链接
链接失效反馈
官方服务:
资源简介:
![11.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/iSho0GesLdh2P9mNr8WOg.png) # **Multilabel-Portrait-18K** **Multilabel-Portrait-18K** is a multi-label portrait classification dataset designed to analyze and categorize different styles of portrait images. It supports classification into the following four portrait types: - **0** — Anime Portrait - **1** — Cartoon Portrait - **2** — Real Portrait - **3** — Sketch Portrait This dataset is ideal for training and evaluating machine learning models in the domain of portrait-style classification. The goal is to enable accurate recognition of artistic and real-world portraits for applications such as image generation, enhancement, style transfer, and content moderation. ## **Use Cases** - Multi-label classification for style recognition - Pretraining or fine-tuning portrait classifiers - Improving filters and sorting in creative AI applications - Enhancing deepfake detection via portrait-style understanding - Style-transfer or portrait enhancement tools ## **Dataset Details** - **Total Samples**: 18,000 portrait images - **Labels**: Multi-label format (each image may have more than one label) - **Label Schema**: - `0`: Anime Portrait [4,444] - `1`: Cartoon Portrait [4,444] - `2`: Real Portrait [4,444] - `3`: Sketch Portrait [4,444] ## **Format** The dataset is typically provided in either: - A directory structure grouped by label - Or a `.csv` / `.json` file containing `filename` and `labels` fields Example (`.csv`): ```csv filename,label portrait_001.jpg,"[0, 3]" portrait_002.jpg,"[2]" portrait_003.jpg,"[1, 2]" ```

![11.png](https://cdn-uploads.huggingface.co/production/uploads/65bb837dbfb878f46c77de4c/iSho0GesLdh2P9mNr8WOg.png) # **多标签肖像数据集18K(Multilabel-Portrait-18K)** **多标签肖像数据集18K(Multilabel-Portrait-18K)** 是一款专为分析与分类不同风格肖像图像而设计的多标签肖像分类数据集,支持对以下四类肖像进行分类: - **0** — 动漫肖像(Anime Portrait) - **1** — 卡通肖像(Cartoon Portrait) - **2** — 写实肖像(Real Portrait) - **3** — 素描肖像(Sketch Portrait) 该数据集非常适合用于肖像风格分类领域内机器学习模型的训练与评估,其目标是实现对艺术肖像与写实肖像的精准识别,可应用于图像生成、图像增强、风格迁移以及内容审核等场景。 ## **应用场景** - 风格识别的多标签分类任务 - 肖像分类器的预训练或微调 - 优化创意人工智能(AI)应用中的滤镜与排序功能 - 通过肖像风格理解提升深度伪造(deepfake)检测能力 - 风格迁移或肖像增强工具 ## **数据集详情** - **总样本量**:18000张肖像图像 - **标签格式**:多标签格式(单张图像可对应多个标签) - **标签规范**: - `0`:动漫肖像(Anime Portrait),样本量4444 - `1`:卡通肖像(Cartoon Portrait),样本量4444 - `2`:写实肖像(Real Portrait),样本量4444 - `3`:素描肖像(Sketch Portrait),样本量4444 ## **数据格式** 该数据集通常以下列两种形式提供: - 按标签分组的目录结构 - 或包含`filename`与`labels`字段的`.csv`或`.json`文件 示例(`.csv`格式): csv filename,label portrait_001.jpg,"[0, 3]" portrait_002.jpg,"[2]" portrait_003.jpg,"[1, 2]"
提供机构:
maas
创建时间:
2025-04-22
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作