A PlastiVision-Edge Enabled Plastic Waste Bank Framework for Community Income Generation and Circular Economy Transition in Rural and Urban India

India generates a large and growing volume of plastic waste each year, but community-level collection stays under-resourced: Self-Help Groups (SHGs) lack an affordable way to identify polymer type at the point of deposit. We present PlastiVision-Edge, a zero-shot, open-vocabulary smartphone detector that identifies plastic waste by material type without labelled examples of every new packaging design. Its core enabler is the Plastic Waste Vocabulary Ontology (PWVO), a 47-category, material-aware prompt set built on Basel Convention waste codes. PWVO prompts a standalone YOLOE-26 model that receives no additional fine-tuning, so recognition comes from the prompt design itself, not from further training. This ontology underpins a proposed Plastic Waste Bank framework involving seven stakeholders. The framework pays citizens per kilogram by polymer type, not a flat mixed-waste rate, and records each transaction as traceability evidence for Extended Producer Responsibility (EPR). On TACO (Trash Annotations in Context, a crowd-annotated litter-photo dataset; 459 images), PWVO-guided YOLOE-26 reaches 8.4% box and 8.3% mask average precision on estimable categories. For 13 packaging types with no labelled examples in any benchmark dataset, it succeeded in 80.0% of trials (95% confidence interval: 65.2–89.5%), tested on multiple author-captured photographs per type. On TrashNet (a studio-photo recyclables dataset), it flags plastic content in 25.7% of images. The model is light enough for field use, running at 41.5 FPS on a GPU and 440 ms per image on Android in a 10.7 MB offline-capable footprint, and its pixel-level output estimates scene-level plastic coverage, unlike box-only detectors. A 436-household urban field pilot shows a 67–108% per-kilogram value uplift over baseline, worth an estimated INR 46–75 per household per month in added income.