BD-Food33: A 1.2K Image Dataset of 33 Bangladeshi Local Food Categories
Description
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.
Files
Steps to reproduce
The dataset was constructed using the following process: Collection of images from: Publicly accessible food photography websites Bangladeshi restaurant menus Food blogs Manually captured photos Filtering for quality (removal of blurred or duplicate images) Cropping and resizing to maintain visual consistency Manual labeling into 33 distinct categories Standardization of filenames and directory structure Creation of metadata CSV including image_id and category Because the dataset includes partially sampled and manually collected images, exact reproduction is not possible, but the final curated dataset is fully provided.
Institutions
- Daffodil International University