Implementation Of SSD MobileNet V2 FPNLite For Website-Based Indonesian Sign Language Recognition

This study implements the SSD MobileNet V2 FPNLite 320x320 model for Indonesian Sign Language (BISINDO) recognition on a web-based application. The main objectives of this study are to determine the time required by the SSD MobileNet V2 FPNLite model for Indonesian Sign Language recognition on a web-based application and to analyze how good the SSD MobileNet V2 FPNLite model is for Indonesian Sign Language recognition. This model offers high efficiency with an average detection time of 0.27 ms per letter, a relatively small model size (7.4 MB), and evaluation results with a mean Average Precision (mAP) of 95.7% and an F1-Score of 93.5%. These results indicate that the SSD MobileNet V2 FPNLite 320x320 is an efficient and accurate solution for web-based BISINDO detection applications.