Herbicide spraying is used to control weeds in the agriculture industry. This technique results in significant waste and pesticide expenditures for farmers and environmental degradation. Allocating the proper herbicide dosages in the proper locations and times is one technique to reduce expenses and environmental impact. Unmanned Aerial Vehicles (UAVs) are rapidly gaining recognition as an effective tool for weed detection and management. Their ability to capture high-resolution images of agricultural fields at a low cost has made them a popular choice in precision agriculture. Automatic weed detection is a challenging task despite significant improvements in UAV capture systems because of how closely weeds resemble the crops. Automated weed detection can help eliminate weeds with minimal harm to people and crops. In this system, the Convolutional Neural Network’s deep learning method is used for the detection of weeds through UAV-captured photos. The suggested approach consists of four key functions. Constructing and using UAV to capture images of the farm, training and testing with CNN to detect the weeds, displaying the results via a web application, and automatically spraying the required quantity of pesticide on weeds. This paper is focused on the construction of UAVs, modeling CNN for training and testing the weeds data, detecting weeds, analyzing the accuracy of prediction, and spraying the pesticide on the weeds through UAV.