In the absence of an effective trash management framework, trash finds its way into canals and drains. Trash objects, particularly plastic waste, are less prone to natural decomposition and often find their way to rivers and eventually reach the agricultural fields. Plastic waste decomposes into micro-plastics, which are produced through the combination of hazardous chemicals, adversely affect the microorganisms in the agricultural soil. The existing techniques that allow for trash classification in water bodies like canals and storm drains are only limited to trash/non-trash or plastic/non-plastic classification and localization. This is mainly because of challenges such as appearance deformation and significant occlusion. For the first time, in this work, we propose a novel 12 class trash classification dataset (TCD) for this problem. A novel attention layer is proposed to solve the problem of trash classification and localization. We also propose a 14 class video dataset called trash flow rate dataset (TFD) for trash class distribution analysis along with a water flow calculation method based on visual sensors. Log attention module was successfully employed in the architectures of state-of-the-art models for better classification, localization, and detection on the trash classification dataset (TCD). An increase in 10% mAP was observed with the use of the log attention module proposed in this thesis. Study conducted on the trash flowrate dataset (TFD) revealed plastic wrapper to be a major cause of plastic pollution regardless of the locality of water channels. Water flow calculation was performed using orthorectification through homography for all sites in the trash flow rate dataset(TFD).