Deep learning for daily care: medicine recognition and reminder systems for the visually impaired

Abstract This research uses a deep learning-based approach that is helpful for the visually impaired humans in their medications, regular life activities without any support from other normal humans/nurses/caretakers. In this study, the inception model is introduced and is trained by 4000 images of the targeted medicines as follows: Dolo, Strepsils, Jovees shampoo, and Volini. The purpose of using many images in this study is to train the model with the training images of the different insights, variations in the distance, and inclination of medicines. After many experimentations of model training with tuning their hyperparameters, the model is trained for 98% training accuracy and 96.98% validation accuracy, which highlights the effectiveness of the system. To create an interactive UI for demonstration, the purpose of the local host is created, where pressing any key on the keyboard activates the camera input. The user is supposed to press any key on keyboard and keep the medicines in front of camera in a random way. The camera will capture the image and take it as input frame for processing purpose through the inception model. After processing the audio, output information will be announced through speaker about the medicine uses and doctor prescription for that particular medicines. The extra facility is also merged in this study to add on the timing and the name of the medicine by nurses that is to be helpful for the blind persons. The uses of the inception model for the medicine recognition and management create the awareness about the advanced program utilization to make easy for the visually impaired persons. All the medicines used in this study are trademark company products of respective owners. The products used in this study are only for research purposes and does not imply any endorsement or affiliation.