Abstract: The widespread presence of micro-plastics in waterbodies lead to great impact on aquatic organisms and our ecosystem. Now a days, water pollution is increasing day by day through human inference. So we need to reduce it by employing some techniques. Traditional methods to detect the micro- plastics are time consuming, labor intensive and limited in accuracy. Recent advancements in deep learning provides promising solutions to automate and enhance the detection and classification of micro- plastics. This study proposes the deep learning models (different types of CNN) such as U-Net or ResNet-50 or Faster R-CNN to recognize the micro-plastic content in different waterbodies. Here the proposed model utilizes a mobile camera to capture the images or upload any images to identify the micro-plastics based on chemical composition. These models are trained with the dataset (micro-plastic dataset for computer vision) ensures robustness in distinguishing micro-plastics from other particles. U-Net provides the best accuracy compare with other models.