Non-Biodegradable Plastic Waste Detection and Classification Using Deep Learning: Bangladeshi Environmental Scenario

Non-biodegradable plastic waste is a major problem in Bangladesh. In particular, YOLOv8 is the most popular deep-learning technique for object recognition which is utilized for waste detection. Detecting and classifying some of the most common waste products on streets, yards, marketplaces, parks, educational institutions, etc. can be aided by VGG16, AlexNet, and ResNet50 which are used for higher validation accuracy. The Non-Biodegradable Plastic Waste Detection and Classification (NPWDC) method begins with an input picture that is fed into the YOLOv8m. It then outputs a cropped piece of the detected region along with a bounding box surrounding the waste area that has been discovered. The resulting picture will now be fed into many classification methods, such as ResNet50, VGG16, and AlexNet, which produce categorized output labeled with the image class name. Since this is an extremely difficult task, the original NPWDC dataset which was created by taking pictures of the wastes under various situations and from various angles is used to train the NPWDC model. This dataset is based on the actual environmental scenario in Bangladesh. The total size of the dataset after augmentation is 10024. According to test data, the NPWDC model has an impressive accuracy rate of 93%. With a focus on deep learning's ability to solve non-biodegradable plastic trash, the NPWDC highlights how deep learning may help Bangladesh and similar regions have a cleaner, more sustainable future.