Image Dataset of Street Food Hygiene of Dhaka

The dataset comprises 1000 annotated images of street food vendors operating in various neighborhoods of Dhaka, Bangladesh. Images were captured using high-resolution smartphone cameras during daytime hours, ensuring optimal lighting conditions for accurate annotation. Each image is categorized based on five hygiene indicators: glove usage (yes/no), food covering (covered/uncovered), surface cleanliness (clean/dirty), waste presence (present/absent), and vendor appearance (clean/unclean). These labels are included in accompanying JSON and CSV annotation files, providing compatibility with popular machine learning frameworks such as TensorFlow and PyTorch. The dataset is structured into folders labeled by hygiene score and vendor ID to maintain clarity. Image resolution ranges between 1080x720 and 1920x1080 pixels. Most images were captured from a frontal or slightly elevated angle to provide a clear view of both the vendor and the food items. The annotation process involved three reviewers who independently labeled each image and resolved discrepancies through consensus. The dataset was preprocessed to exclude blurry or obstructed images, ensuring a high-quality resource for training and benchmarking. A distribution chart of hygiene indicators across different neighborhoods is included in the metadata, offering additional insights into geographic hygiene patterns within the city.