We present the system that we have developed for the identification and\nverification of pills using images that are taken by the VeriMedi device. The\nVeriMedi device is an Internet of Things device that takes pictures of a filled\npill vial from the bottom of the vial and uses the solution that is presented\nin this research to identify the pills in the vials. The solution has two\nserially connected deep learning solutions which do segmentation and\nidentification. The segmentation solution creates the masks for each pill in\nthe vial image by using the Mask R-CNN model, then segments and crops the pills\nand blurs the background. After that, the segmented pill images are sent to the\nidentification solution where a Deep Metric Learning model that is trained with\nProxy Anchor Loss (PAL) function generates embedding vectors for each pill\nimage. The generated embedding vectors are fed into a one-layer fully connected\nnetwork that is trained with the exact solution to predict each single pill\nimage. Then, the aggregation/verification function aggregates the multiple\npredictions coming from multiple single pill images and verifies the\ncorrectness of the final prediction with respect to predefined rules. Besides,\nwe enhanced the PAL with a better proxy initialization that increased the\nperformance of the models and let the model learn the new classes of images\ncontinually without retraining the model with the whole dataset. When the model\nthat is trained with initial classes is retrained only with new classes, the\naccuracy of the model increases for both old and new classes. The\nidentification solution that we have presented in this research can also be\nreused for other problem domains which require continual learning and/or\nFine-Grained Visual Categorization.\n