Real-Time Embedded Vision System for Plastic Bag Detection on Raspberry Pi 5

Plastic bags have become one of the most troublesome forms of waste because they are cheap, used everywhere, and do not break down easily. Once thrown away, they often end up blocking drains, floating in rivers, or scattered in natural spaces, causing serious problems for both the environment and infrastructure. Sorting this waste by hand is slow and unreliable, while large automated systems are usually too expensive for small communities or field applications. This work presents a simple and affordable system that can detect plastic bags using a Raspberry Pi 5 and a lightweight object detection model. The model was trained with a custom dataset of 2,909 labeled images through Roboflow and deployed using YOLOv11. The device captures video at a modest resolution of 320×240 pixels and 10 frames per second. Each frame is sent through a cloud-based API for analysis, and the results are improved with extra image filtering to cut down on false alarms. Whenever a bag is detected, the system sends a push notification with the image and details to the user. Tests showed strong results, with the model reaching 99.0% mAP, 99.4% precision, and 97.8% recall. Overall, the design is practical, low-cost, and flexible enough for real-world waste monitoring.