Smartphone-based multi-criteria vegetable object detection dataset from Bangladesh.

The agricultural landscape of Bangladesh is significantly influenced by the cultivation of vegetables, which is directly involved in the nutritional intake of the people, the economy, and the food security of the entire nation. Precise identification of vegetables is essential to efficient cultivation, inventory management, and smart agricultural practices. In our study, we introduce a comprehensive dataset for vegetable detection, consisting of 3534 high-resolution images captured in natural, real-world settings using a Redmi Note 12 from multiple roadside vehicles of local vendors. The dataset encompasses 22 distinct vegetable classes, covering a wide range of appearances, shapes, and natural daylight environments to enhance model robustness and practical applicability. Each image has been meticulously annotated using the Roboflow platform to facilitate object detection tasks, and the resulting dataset is provided in Pascal VOC format. Unlike other imaging datasets, our work emphasizes ground-level perspectives, making it particularly relevant for handheld and low-cost monitoring systems. The primary goal of this dataset is to support the development of computer vision models for accurate vegetable recognition, thereby aiding decision-making in vegetable cultivation and contributing to smarter and sustainable agricultural practices.