Effective weed management becomes increasingly important as the demand for sustainable farming practices increases. Deep Learning (DL), an area of Artificial Intelligence (AI) is emerging as a feasible solution due to its ability to extract complex patterns from large data. This review paper addresses the concerns about weeds in agriculture and how DL techniques, alongside computer vision methods, can be used to detect weeds in agricultural fields. This also compares the traditional weed detection methods with DL approaches, discussing the optimal stage to control the weeds and how Machine Learning (ML) and DL techniques are employed for detection. It compares the performance of several DL architectures, dataset sizes, and training procedures in weed detection tasks. Furthermore, addresses the issues such as dataset scarcity, model interpretability, and real-world deployment, as well as potential remedies and future research initiatives. With this review, the authors have synthesized the existing literature, highlighting relevant findings, helpful to create more accurate and scalable weed detection systems for sustainable agriculture.