Smart Door Automation System with DeepFace and SLMs
- Author Sudip Chakraborty
- Co-Author Deep Chakraborty
- DOI https://ww
- Country : India
- Subject : Computer Science and Information Sciences
The continuous evolution of smart home technologies has underscored the necessity for secure, efficient, and user-friendly solutions in residential automation. Addressing these requirements, we introduce an advanced Smart Door Automation System, innovatively integrating DeepFace facial recognition technology with the compact and effective language model, TinyLLaMA. This integration aims to provide residents with reliable security and personalized interactions, transforming routine home entry into an intuitive, seamless experience. This innovative system leverages standard CCTV camera feeds to accurately and rapidly detect authorized users, thereby significantly strengthening home security without interrupting daily routines. DeepFace, renowned for its robust facial recognition capabilities, adeptly handles variations in environmental lighting, facial orientations, and minor changes in user appearance, consistently ensuring reliable and accurate identification. By conducting both the facial recognition and linguistic processing directly on the local device, our system considerably reduces response times, eliminates potential network latency, and enhances data privacy by minimizing cloud dependency. Upon positive user recognition, TinyLLaMA swiftly generates personalized, context-specific greetings, which are instantly transformed into clear, natural-sounding audio messages through Google Text-to-Speech (GTTS). The combined efficacy of DeepFace and TinyLLaMA has been validated through extensive testing, demonstrating a notable reduction in unauthorized access attempts alongside improved user engagement and satisfaction. Overall, this comprehensive approach illustrates the practical application of advanced artificial intelligence methods, significantly raising the standards of everyday smart home security and automation.
Comments
No have any comment !