Abstract—This study was conducted to determine the optimal time for changing stoma appliances. One of the problems with stoma appliances is that it is difficult to regularly change the faceplate. The faceplate is sticky and is attached around the stoma. If it is used for a long time, urine or feces may seep between the faceplate and the stoma, causing the faceplate to come off and urine or feces to leak. However, patients themselves need to be accustomed to changing stoma appliances. In addition, if nurses who are not used to changing the appliance do the job, there are cases where the timing of the change is mistaken. The timing of changing the stoma appliance is determined by the amount of urine and feces of the patient, the color change of the appliance, and the degree of dissolution. Therefore, it is expected that if the patients themselves change the appliance, it will prevent mistakes in the timing of the change and leakage of urine and feces, through image processing and machine learning. 
 

Keywords—stoma, machine learning, teachable machine, NMKD stable diffusion, grayscale HSV, smoothing, sharpening, Gaussian filter, overlearning