Open renmengyuan opened 4 years ago
You can use the faces with closes eyes to train this model linking is https://github.com/zhangqianhui/GazeCorrection
You can use the faces with closes eyes to train this model linking is https://github.com/zhangqianhui/GazeCorrection
Thanks!The DeepWarp is your baseline model, could you share the implementation of DeepWarp. Very thanks!
hi @zhangqianhui : Do you know how to separate generator model from the only one checkpoint model to inference? The generator model should have only two input: x_test_r and angle_test_g
You can define a new the saver ?
Now I'm trying to use the restore method. Define the scope = 'generator', then restore the model from the ckpt. Finally, save it to a .pb file. However, i'm not sure if it will be the same as the result produced by the original ckpt model.
Do you want to freeze this model ?
Could the model make the eyes closed?