Brazilian_Brand_Marketing_Banners_Dataset

# Brazilian Brand Marketing Banners Dataset
*This dataset contains high-quality images of Brazilian brand marketing banners collected from online and offline retail environments. It includes product advertisements, promotional offers, and digital marketing visuals designed for both Portuguese-speaking audiences and bilingual markets.*

## Contact
For queries or collaborations related to this dataset, contact:
- [e-mail]
- [e-mail]

## Supported Tasks

- **Task Categories**:
- Image Classification
- Text Recognition (OCR)
- Multimodal Understanding

- **Supported Tasks**:
- Brand logo recognition and classification
- Text extraction from marketing visuals
- Sentiment and tone analysis of ad language
- Visual layout analysis of marketing banners
- AI training for ad design optimization and localization
- Multimodal understanding of visual-text advertising content

## Languages

- **Primary Language**: Portuguese
- **Secondary Presence**: English (on bilingual or global brand campaigns)

## Dataset Creation

### Curation Rationale
The dataset was curated to enable research and development of AI systems capable of understanding brand identity, marketing language, and visual design principles in advertising. It supports models in OCR, multimodal learning, and visual-semantic representation for retail and ad-tech applications.

### Source Data
- **Contributors**: Collected from Brazilian e-commerce sites, outdoor advertising spaces, and social media brand pages
- **Collection Process**: Banners were captured or archived from public sources. All personal information or identifiable human subjects were excluded.

### Other Known Limitations
- **Bias**: Prominent national and international brands are more represented than small local businesses
- **Image Quality**: Some banners may have compression artifacts or uneven lighting conditions
- **Temporal Relevance**: Ads reflect specific promotional periods and may not represent current marketing trends

## Intended Uses

### ✅ Direct Use
- Training OCR and visual-text models for marketing content
- Brand and logo recognition research
- Multimodal AI for understanding ad layouts and brand tone
- Retail and marketing analytics
- Cross-lingual advertisement localization

### ❌ Out-of-Scope Use
- Commercial reproduction of copyrighted brand materials
- Misuse of ads for misleading or defamatory purposes
- Creation of synthetic promotional content mimicking real brands without consent

## License

CC BY 4.0

# 巴西品牌营销横幅数据集
本数据集收录了从线上及线下零售场景中采集的高质量巴西品牌营销横幅图像，涵盖面向葡萄牙语受众及双语市场设计的商品广告、促销活动物料与数字营销视觉素材。

## 联系方式
若您对此数据集存在查询或合作意向，请通过以下邮箱联系：
- [e-mail]
- [e-mail]

## 支持任务
- **任务类别**：
- 图像分类
- 文本识别（光学字符识别，OCR）
- 多模态理解
- **支持任务**：
- 品牌标识（Brand Logo）识别与分类
- 营销视觉素材文本提取
- 广告语言的情感与语气分析
- 营销横幅视觉布局分析
- 面向广告设计优化与本土化的AI训练
- 视觉-文本广告内容的多模态理解

## 语言覆盖
- **主要语言**：葡萄牙语
- **次要语言覆盖**：针对双语或全球品牌营销活动的素材中包含英语

## 数据集构建
### 采集逻辑
本数据集的采集旨在推动可理解广告中品牌标识、营销语言与视觉设计原则的AI系统研发，可服务于零售与广告技术领域中光学字符识别（Optical Character Recognition，OCR）、多模态学习及视觉语义表征相关的模型训练。

### 源数据
- **采集渠道**：数据采集自巴西电商平台、户外广告点位及社交媒体品牌主页
- **采集流程**：横幅素材从公开渠道捕获或归档，所有个人信息及可识别的人类主体均已剔除。

### 已知局限性
- **偏倚性**：相较于小型本地商户，头部本土及国际品牌的素材占比更高
- **图像质量**：部分横幅可能存在压缩伪影或光照不均的问题
- **时效性**：广告素材反映特定促销时段的内容，未必契合当前营销趋势

## 预期用途
### ✅ 直接适用场景
- 营销内容的光学字符识别与视觉-文本模型训练
- 品牌与标识识别相关研究
- 理解广告布局与品牌语气的多模态AI研发
- 零售与营销分析
- 跨语言广告本土化

### ❌ 禁用场景
- 对受版权保护的品牌素材进行商业复刻
- 误用广告素材进行误导性或诽谤性活动
- 未经授权制作模仿真实品牌的合成促销内容

## 授权协议
CC BY 4.0