This thesis presents the design and implementation of a Raspberry Pi 5–based smart lock that uses CNN powered facial recognition with RFID as a fallback method. It explains the theory behind modern face recognition, including detection, alignment, embedding, matching, and software anti-spoofing to defend against printed, screen, and video attacks. The system integrates a camera, servo lock, RFID module, and custom PCB with power management to create a complete embedded hardware platform. A full software pipeline using YuNet, SFace, cosine similarity, and DeepFace liveness detection provides secure user verification. The prototype is also connected to a mobile app via MQTT for remote control and monitoring. Testing shows that the device reliably identifies enrolled users, rejects impostors, and demonstrates that low-cost 2D vision can provide secure home access.