KwaiVGI
请问是怎么定量计算表情和隐式特征点之间的关系的 How to quantify expressions and implicit lmks
It appears that we didn’t provide a quantification for it. You can refer to LivePortrait paper for more details.
我不是很理解你们是怎么在app.py文件里面对特征点进行的定量操作，比如说smile调整的lmk以及wink调整的lmk。 In your app.py, how do you quantitatively calculate the relationship between SMILE and LMKS
我不是很理解你们是怎么在app.py文件里面对特征点进行的定量操作，比如说smile调整的lmk以及wink调整的lmk。 In your app.py, how do you quantitatively calculate the relationship between SMILE and LMKS
一样的问题。在代码中还出现了： for idx in [1,2,6,11,12,13,14,15,16,17,18,19,20]: delta_new[:, idx, :] = x_d_exp_lst_smooth[i][idx, :] if flag_is_source_video else x_d_i_info['exp'][:, idx, :] delta_new[:, 3:5, 1] = x_d_exp_lst_smooth[i][3:5, 1] if flag_is_source_video else x_d_i_info['exp'][:, 3:5, 1] delta_new[:, 5, 2] = x_d_exp_lst_smooth[i][5, 2] if flag_is_source_video else x_d_i_info['exp'][:, 5, 2] delta_new[:, 8, 2] = x_d_exp_lst_smooth[i][8, 2] if flag_is_source_video else x_d_i_info['exp'][:, 8, 2] delta_new[:, 9, 1:] = x_d_exp_lst_smooth[i][9, 1:] if flag_is_source_video else x_d_i_info['exp'][:, 9, 1:]不知道这些点是如何设定的
我不是很理解你们是怎么在app.py文件里面对特征点进行的定量操作，比如说smile调整的lmk以及wink调整的lmk。 In your app.py, how do you quantitatively calculate the relationship between SMILE and LMKS
一样的问题。在代码中还出现了：
for idx in [1,2,6,11,12,13,14,15,16,17,18,19,20]: delta_new[:, idx, :] = x_d_exp_lst_smooth[i][idx, :] if flag_is_source_video else x_d_i_info['exp'][:, idx, :] delta_new[:, 3:5, 1] = x_d_exp_lst_smooth[i][3:5, 1] if flag_is_source_video else x_d_i_info['exp'][:, 3:5, 1] delta_new[:, 5, 2] = x_d_exp_lst_smooth[i][5, 2] if flag_is_source_video else x_d_i_info['exp'][:, 5, 2] delta_new[:, 8, 2] = x_d_exp_lst_smooth[i][8, 2] if flag_is_source_video else x_d_i_info['exp'][:, 8, 2] delta_new[:, 9, 1:] = x_d_exp_lst_smooth[i][9, 1:] if flag_is_source_video else x_d_i_info['exp'][:, 9, 1:]不知道这些点是如何设定的
同样的问题
