An Indian annotated weed dataset for computer vision tasks in precision farming.

Weed infestations are the major threat for agriculture sector in India, significantly impacting crop productivity. These invasive plants not only attract pests but also compete with crops for essential nutrients, contributing to an estimated 45 % of the annual productivity loss in agriculture. For smallholder farmers, traditional methods such as manual weeding is both labour-intensive and expensive. Heavy reliance on usage of chemical herbicides has led to resistance in several weed species. Emerging technologies such as artificial intelligence and computer vision are transitioning farming sector by automating tasks. The main component for development of these technologies is the availability of datasets. To address this need, a comprehensive MH-Weed16 image dataset is created which consists of total 25,972 images acquired from real fields of Maharashtra region. Dataset includes 16 different weed species, annotated under guidance of agriculture experts. Out of total, dataset contains 7577 samples featuring both crops and weeds, captured from a top view to ensure precise estimation of weed areas. The proposed dataset will serve as a valuable resource for computer vision tasks in precision farming. The objective of this research is to contribute towards integrating technology for weed management strategies, paving the way for sustainable agricultural practices.