AGE DETECTION IN REAL TIME VIDEOS USING DEEP LEARNING
Keywords:
Deep learning, Real-Time Age Detection, Age identificationAbstract
Knowing the age of humans is becoming more relevant in many aspects of our lives and its used differently. Inaccuracy in age detection can sometimes be a limitation or detrimental. Age is an important unmodifiable human trait that is even used as a
deciding factor for employment, marketing, investigation, identification or for other organizational protocols. Current systems lack the required versatility, relevance, speed and efficiency to cope with the changing IR 4.0 industrial needs. This research
presents a fast, efficient and user-friendly system that is able to detect human age in real time videos using deep learning. The proposed system can be used in offices, companies, stores, mosques, and other places. In this study, a qualitative method was
applied as the research method. Open ended interviews were conducted to gather the user requirements. The findings from the interviews proved features like gender detection and age detection in static images are no longer needed by consumers. The proposed system uses the Haar Cascades algorithm to identify human faces from real time video frames. These frames are processed by an age detector model. The results of this computation gives an output of an accurate age range. The proposed system can detect and predict the age of multiple faces in a single video frame, allowing batch processing. The captured age ranges are saved in a CSV file that can be transformed to a local database. The testing results proved that the proposed system works better in a proper lightening conditions where the face can clearly be seen by the camera. Using the Haar Cascades algorithm proved that age can be predicted even with sunglasses, with the eyes closed or virtually from a smartphone video. The proposed system will help end users eliminate errors and manual work in age detection for their various needs. Efficiency will be increased through batch processing. Conclusively, the objectives of this research have been successfully achieved.
