Indonesian Lip-Reading Detection and Recognition Based on Lip Shape Using Face Mesh and Long-Term Recurrent Convolutional Network

Table 6

Comparison of the results of the proposed method with previous work in MIRACL-VC1.

Recognition methods	Detection methods	Dataset	Accuracy (%)	F1 score (%)
Fine-tune VGG-16 + LSTM [47]	Dlib	MIRACL-VC1	59	Unspecified
Inception V3 [48]	Dlib	MIRACL-VC1	86.6	Unspecified
CNN + Batch normalization [49]	Haar cascade	MIRACL-VC1	52.9	Unspecified
MobileNet + LSTM [50]	HOG + SVM	Modified MIRACL-VC1	94	84
VGG-16 + LSTM [50]	HOG + SVM	Modified MIRACL-VC1	96	70
3D-CNN [51]	Dlib	MIRACL-VC1	70.2	Unspecified
LSTM [52]	Dlib	MIRACL-VC1	66	66
3D-CNN + LSTM [52]	Dlib	MIRACL-VC1	85	Unspecified
LRCN-2Conv	MediaPipe	MIRACL-VC1	81.33	81
LRCN-3Conv	MediaPipe	MIRACL-VC1	90.67	91