search

wuhaozhe / style_avatar

A repository for generating stylized talking 3D and 3D face
278 stars 34 forks source link

bad result #6

Closed liyuanyaun closed 3 years ago

liyuanyaun commented 3 years ago

use your model, demo code, example.png and example.wav, but the generated result is bad 捕获

what is the problem？ @wuhaozhe

wuhaozhe commented 3 years ago

image hi, here is my generation results, let's check whether the model weight is correct?

could you please print the following results?

print(torch.load("./render/model/face_unet.pkl")['model.13.conv_block.1.weight'][0][0])
print(torch.load("./render/model/tex_encoder.pkl")['down1.conv.0.weight'][0][0])

My results are

tensor([[ 0.2953,  0.0757, -0.0737],
        [-0.1960, -0.0517,  0.0119],
        [ 0.0682,  0.2076,  0.0557]], device='cuda:0')

tensor([[ 0.1049, -0.1246,  0.0850],
        [-0.4058, -0.0189,  0.0529],
        [-0.1202, -0.0470, -0.0942]], device='cuda:0')
liyuanyaun commented 3 years ago

image hi, here is my generation results, let's check whether the model weight is correct?

could you please print the following results?

print(torch.load("./render/model/face_unet.pkl")['model.13.conv_block.1.weight'][0][0])
print(torch.load("./render/model/tex_encoder.pkl")['down1.conv.0.weight'][0][0])

My results are

tensor([[ 0.2953,  0.0757, -0.0737],
        [-0.1960, -0.0517,  0.0119],
        [ 0.0682,  0.2076,  0.0557]], device='cuda:0')

tensor([[ 0.1049, -0.1246,  0.0850],
        [-0.4058, -0.0189,  0.0529],
        [-0.1202, -0.0470, -0.0942]], device='cuda:0')

i try to print, but face_unet.pkl is different from you 捕获

wuhaozhe commented 3 years ago

image hi, here is my generation results, let's check whether the model weight is correct? could you please print the following results?

print(torch.load("./render/model/face_unet.pkl")['model.13.conv_block.1.weight'][0][0])
print(torch.load("./render/model/tex_encoder.pkl")['down1.conv.0.weight'][0][0])

My results are

tensor([[ 0.2953,  0.0757, -0.0737],
        [-0.1960, -0.0517,  0.0119],
        [ 0.0682,  0.2076,  0.0557]], device='cuda:0')

tensor([[ 0.1049, -0.1246,  0.0850],
        [-0.4058, -0.0189,  0.0529],
        [-0.1202, -0.0470, -0.0942]], device='cuda:0')

i try to print, but face_unet.pkl is different from you 捕获

My be something wrong with the face_unet.pkl, I have updated the download link, you can try again.

liyuanyaun commented 3 years ago

image hi, here is my generation results, let's check whether the model weight is correct? could you please print the following results?

print(torch.load("./render/model/face_unet.pkl")['model.13.conv_block.1.weight'][0][0])
print(torch.load("./render/model/tex_encoder.pkl")['down1.conv.0.weight'][0][0])

My results are

tensor([[ 0.2953,  0.0757, -0.0737],
        [-0.1960, -0.0517,  0.0119],
        [ 0.0682,  0.2076,  0.0557]], device='cuda:0')

tensor([[ 0.1049, -0.1246,  0.0850],
        [-0.4058, -0.0189,  0.0529],
        [-0.1202, -0.0470, -0.0942]], device='cuda:0')

i try to print, but face_unet.pkl is different from you 捕获

My be something wrong with the face_unet.pkl, I have updated the download link, you can try again.

this solved my case, thank you