A Novel CNN-Based Approach for Intelligent Garbage Detection and Localization

In India, the problem of waste management is one of the most critical concern as tonnes of waste is generated every year of which only a very minute percentage is treated. The rest of it is either are not treated properly or dumped in landfills. Governments around the world have extensively tried to develop intelligent waste or garbage management techniques. The potential threat to sanitation, public health and safety caused by improper waste management is immense. In light of rapid increase in computational efficiency as neural networks became more capable of processing deeper data, deep learning was inevitable for garbage management systems. This paper attempts to provide an end-to-end solution to automating the process of garbage detection and alerts, beginning right at the source. The solution proposed comprises employing a highly accurate training strategy using the acclaimed VGG-16 network using a subset of the COCO dataset and also includes an accompanying Android app that makes the process of garbage localization and alerts seamless and effective. Finally, we guide you through our mobile application that is designed to assist municipalities in remotely monitoring garbage accumulation areas which help to take appropriate measures to keep surrounding areas clean and hygienic. The experimental accuracy of this system was over 95.92% which is much greater than that required for real-world scenarios.