search

yfeng95 / face3d

Python tools for 3D face: 3DMM, Mesh processing(transform, camera, light, render), 3D face representations.
2.63k stars 609 forks source link

what if I use another 3DMM model such as BFM2017 ? #37

Open VincentXWD opened 5 years ago

VincentXWD commented 5 years ago

Hi, The model comes from https://gravis.dmi.unibas.ch/PMM/ could you plz have a look at this model ?

HectorAnadon commented 5 years ago

I would be also interested in this model. Especially the UV map and the IBUG landmarks correspondence to vertices. Thank you for your awesome work!

FelixFox commented 5 years ago

I'm using BFM2017 for training. Now it works (uv map for almost every example in the training set is generated, the training process is launched). Later I may answer about its results

VincentXWD commented 5 years ago

There's a well-trained model from http://gravis.dmi.unibas.ch/PMM/. I'm using this model to replace BFM09 for inference.

FelixFox commented 5 years ago

@VincentXWD It's Scalismo (from here ) or it's The Basel Illumination Prior (from here ) ?

VincentXWD commented 5 years ago

I use the BFM17 model and rewrote the IO. The model is from http://gravis.dmi.unibas.ch/PMM/data/overview/

BenjBarral commented 5 years ago

I would be also interested in this model. Especially the UV map and the IBUG landmarks correspondence to vertices. Thank you for your awesome work!

Hi, I am also interested in the BFM2017 model, and specifically the UV parametrization ! Has any body found a way to get it ? Thank you @YadiraF !

BernhardEgger commented 4 years ago

Hi, a UV-parametrization for the BFM2017 can be found here: face mask: https://github.com/unibas-gravis/parametric-face-image-generator/blob/master/data/regions/face12.json bfm mask: https://github.com/unibas-gravis/parametric-face-image-generator/blob/master/data/regions/bfm09.png

ReverseSystem001 commented 4 years ago

@FelixFox the training set used in 3DDFA includes .mat file(GT, which is necessary to the scipy 8_generate_posmap_300WLP.py ). as i now the mat file includes shape_paras , exp_paras and so on. These parameters are generated based on BFM2009. Before you use BFM2017 for training, do you generate these parameters which is based on BFM2017?　Look forward to your reply. thks

nsalminen commented 4 years ago

I use the BFM17 model and rewrote the IO. The model is from http://gravis.dmi.unibas.ch/PMM/data/overview/

@VincentXWD Would you be willing to share your version of the IO for BFM 2017? I would greatly appreciate it if you could!

VincentXWD commented 4 years ago

I use the BFM17 model and rewrote the IO. The model is from http://gravis.dmi.unibas.ch/PMM/data/overview/

@VincentXWD Would you be willing to share your version of the IO for BFM 2017? I would greatly appreciate it if you could!

Sorry I cannot. :-( I was left the company and the modified scripts were saved in there. Currently I don't have permission to access the codes.

nsalminen commented 4 years ago

I use the BFM17 model and rewrote the IO. The model is from http://gravis.dmi.unibas.ch/PMM/data/overview/

@VincentXWD Would you be willing to share your version of the IO for BFM 2017? I would greatly appreciate it if you could!

Sorry I cannot. :-( I was left the company and the modified scripts were saved in there. Currently I don't have permission to access the codes.

Ah okay makes sense! Thanks for your quick reply!