Peritoneal dialysis (PD) is an alternative treatment of home-based dialysis for Acute Kidney Injury (AKI). A patient who undergoes PD requires self-management skills to deliver and manage dialysis at home effectively. The effluent dialysate of PD patients can be an indicator for early diagnosis complications in Peritoneal dialysis. Under normal circumstances, the appearance of effluent dialysate is transparent to yellowish clear. Changes in turbidity level or color of effluent dialysate indicate an extra or intraperitoneal abnormality. This study aims to develop an image processing model to detect and classify effluent dialysate of PD patients for early warning of complications.The model was built into the mobile application system. The dataset was obtained from patients, and secondary data were used. Image augmentation is used to enhance the quantity of data by nine times. The data set were divided for train and validation within the 8:2 ratio. The result shows that our system has a high accuracy of 94.7% and minimal loss.