BD-Food33: A 1.2K Image Dataset of 33 Bangladeshi Local Food Categories

BD-Food33 is a curated image dataset containing 1,200+ high-quality images of 33 categories of Bangladeshi local food. The dataset was collected from multiple publicly accessible online sources, restaurant websites, food blogs, and manually captured samples. Only selected portions from each source were used to ensure diversity and visual clarity.

All images were cleaned, resized, checked for quality, and organized into category-specific folders. The dataset provides a representative visual catalog of Bangladeshi cuisine—covering traditional, regional, snacks, sweets, and main dishes.

This dataset is suitable for research in computer vision, image classification, food recognition, fine-grained visual categorization, and deep learning–based food analysis. It can be used for training convolutional neural networks, benchmarking image recognition models, building food-recommendation systems, and developing automated nutrition/food-identification pipelines.

BD-Food33是一款精心遴选的图像数据集，涵盖33类孟加拉本土美食，共计1200余张高质量图像。该数据集的采集来源包括多个公开在线平台、餐厅官网与美食博客，同时包含手动拍摄的样本。为确保数据集的多样性与画面清晰度，仅从各来源中筛选合适的图像进行使用。
所有图像均经过清洗、尺寸调整、质量校验，并按类别分文件夹整理归档。该数据集提供了具有代表性的孟加拉美食视觉图鉴，覆盖传统美食、地域特色菜品、小吃、甜点与正餐主菜。
该数据集适用于计算机视觉（Computer Vision）、图像分类、食物识别、细粒度视觉分类以及基于深度学习的食物分析等方向的研究。其可用于训练卷积神经网络（Convolutional Neural Networks）、对图像识别模型开展基准评测、搭建食物推荐系统，以及开发自动化营养分析与食物识别流水线。