iVS-FaceID is an application about continual learning into the Facial accessing control system. For traditional facial recognition learning, it is necessary to retrain the AI model repeatedly after mixing the old and new images to recognize the new person. When the number of recognized people accumulates, the time spent on retraining is also increasing. However, adopting continual learning only takes the time of new learners. Compared with traditional joint learning, with the accumulation of retraining times, the benefits show exponential differences.
iVS-FaceID為將持續學習應用於門禁系統，傳統臉部辨識學習若要新增加辨識人員，須將新舊圖資混合後反覆重新訓練AI模型，當人數日積月累，所耗費的再訓練時間也日漸增加。而採用學新不忘舊模型僅需耗費新學人員的時間，與傳統joint learning相較，隨著再訓練次數的累積，其效益呈現指數型差異。