Multi images inferencing about deep3dfacerecon_pytorch

[YuDeng]

Hi, you can take the average of the 'id' coefficients predicted by different images here:

Deep3DFaceRecon_pytorch/models/bfm.py

Line 283 in 73d4911

coef_dict = self.split_coeff(coeffs)

, and set the 'exp' coef to zero. Then you just need to send the modified coef_dict to the reconstruction pipeline and it will return you an average face shape.

[icedpanda]

Correct me if I am wrong. I suspect the model would also need face landmarks from those additional images. If so, this may not apply to my case since other images were captured from the side and do not have landmarks like eyes, nose, etc.

Thanks

[pyxploiter]

Correct me if I am wrong. I suspect the model would also need face landmarks from those additional images. If so, this may not apply to my case since other images were captured from the side and do not have landmarks like eyes, nose, etc.

Thanks

Hi @icedpanda, were you able to run this for multi images?

[icedpanda]

@pyxploiter Haven't had a chance to run multi images yet, but a single image looks fine

[WildPant]

@icedpanda

@pyxploiter还没有机会运行多张图片，但单张图片看起来不错
Have you found a job that fits your dataset?