Virtual Gold Chain Try-on using Augmented Reality with Neck Occlusion Detector

The inability to try-on jewellery continues to limit customer trust in online jewellery shopping. This paper presents a virtual gold chain try-on system using Augmented Reality in real-time. The primary objective is achieving precise neck-based alignment and improving user confidence and interaction. The proposed research work integrates MediaPipe FaceMesh for landmark detection, neck occlusion detector for detecting obstructions caused by hair, clothing, hands and OpenCV for chain overlay, deployed through Streamlit interface. The occlusion model employs CNN classifier MobileNetV2 for feature extraction from the custom dataset which identifies whether the neck is visible or occluded. Evaluation on the 1,200-image dataset resulted in an overall classification accuracy of 92.5%. Experimental evaluation showed that the system maintained a frame rate of 30 FPS with an end-to-end latency of 150 ms, thereby supporting smooth real-time user interaction. This highlights the potential of enhanced user experience, reduces the return rates and can serve as a foundation for other AR applications in retail.