KAIST Lip reading module for HumanCare Project
Test on Ubuntu 16.04, Python3.5.2
dlib==19.6
llvm==8.0
librosa==0.7.2
opencv-python==4.2.0.34
imutils==0.5.3
tensorflow-gpu==1.12.0
-
Download the weight files from HERE and move it in weight folder 1-1. Download the dataset files for the evaluation from HERE
-
Install dlib(install guide: HERE])
-
Install llvm8 and clang8(install guide: HERE)
-
Install the requirements.txt
pip install -r requirements.txt
-
Lipreading Demo Target video has to be placed in "input_avi" folder.
python demo_main.py
-
Lipreading evaluation Evaluation by precision from the dataset/filter3_old2 folder.(Test file should contain the label name in the filename(ex 001_old_d_get1_1.avi) Input file is based on the cropped lip video.
python eval_demo.py