The rapid growth of water weeds and invasive species such as water hyacinth poses a significant threat to aquatic ecosystems and human activities. Early detection and monitoring of these aquatic plants are essential for effective management and mitigation of their detrimental effects. In this paper, we proposition an approach designed for water weed and hyacinth detection using MobileNetV2 as the base network and Feature Pyramid Network (FPN) Lite with an input size of $320\times 320$ pi of TensorFlow version. We created the dataset of Pune city localize area water weeds and hyacinth images. The model is trained using a large dataset of annotated images containing various instances of water weed and hyacinth. The dataset includes diverse environmental conditions, illumination variations, and cluttered backgrounds, making the model robust and adaptable to real-world scenarios. The accuracy of the model is about 71.25%, which detects the proper and accurate prediction.