The swift rise in plastic waste buildup along roadways has turned into a major issue because of its ecological risks and detrimental effects on urban tidiness. Traditional techniques for identifying and monitoring plastic waste are primarily manual, leading to inefficiencies, high labor demands, and susceptibility to errors. This research introduces a smart image-based detection system employing Convolutional Neural Networks (CNN) for the automated recognition of plastic waste along roadways. The system records live images using cameras placed on roadside poles, vehicles, or drones. These images undergo a series of preprocessing steps including conversion to grayscale, removal of noise, resizing, and enhancement of edges utilizing OpenCV to get them ready for precise feature learning. A tailored CNN model is subsequently utilized to derive spatial features and accurately classify the occurrence of plastic waste. The CNN is trained on a labeled dataset that includes images featuring both plastic debris and clean backgrounds across different lighting, background noise, and environmental settings. The system is designed for efficient performance, allowing it to be deployed on low-resource edge devices for practical application in the field. Performance assessment shows that the model attains significant accuracy and recall, showcasing its strength in identifying plastic even in situations that are partially obstructed or in low-light conditions. Through the automation of plastic waste detection along roadways, this system offers a scalable and budget-friendly answer for supporting municipal authorities in preserving environmental cleanliness and starting prompt waste management efforts. It additionally creates opportunities for real-time notification systems and smart city connectivity.