The management of plastic waste is complex process worldwide. The manual process to sort garbage is termed to be complex and expensive task and hence the scientists studied automatic sorting technique which can elevate the efficacy of recycling task. Moreover, the waste segregation models are adapted to several groups of materials to sort the waste. The usage of deep models is considered to be effective in classifying the thermoplastic waste segregation. The images are attained and delivered to pre-processing with median filter followed by segmentation. PSI-Net considering Archimedes Henry Gas Solubility Optimization (AHGSO) is adapted to segment pre-processed images. Thereafter, texture features along with statistical features are obtained for undergoing first level classification. The first level categorization is executed using Deep Quantum Neural Network (DQNN) wherein the model is trained using AHGSO and it classifies into three classes, namely “no plastic,”“Resin,” and “plastic.” In addition, the second level classification is done in case of plastic or resin using Deep Maxout Network (DMN)-AHGSO wherein DMN undergoes training by means of AHGSO. The DMN-AHGSO gained premium accuracy of 92.9%, F-measure of 92.6%, and precision of 92.9%.