I found that in your work, facial expressions are generated by specifying numerical values. I tried to transfer the expression boundaries of IdInvert, but I found that it seemed to have no effect. In IdInvert, the dimension of the emoticon boundary is [1,14,512]. I noticed that in your boundary file, all attribute dimensions are [1,512].I tried to integrate the dimensions of [1,14,512] into the form of [1,512] according to the second dimension, but I found that after integration, the expression did not change. The code is as follows:
expression_boundary.shape#[1,14,512] manipulate_layers=[2,3,4,5] expression=np.mean(expression_boundary[:manipulate_layers,:],exis=1) np.save("expression.npy",expression)
I found that in your work, facial expressions are generated by specifying numerical values. I tried to transfer the expression boundaries of IdInvert, but I found that it seemed to have no effect. In IdInvert, the dimension of the emoticon boundary is [1,14,512]. I noticed that in your boundary file, all attribute dimensions are [1,512].I tried to integrate the dimensions of [1,14,512] into the form of [1,512] according to the second dimension, but I found that after integration, the expression did not change. The code is as follows:
expression_boundary.shape#[1,14,512] manipulate_layers=[2,3,4,5] expression=np.mean(expression_boundary[:manipulate_layers,:],exis=1) np.save("expression.npy",expression)