chahuja / mix-stage Goto Github PK

Hello ! Can it be appliable for the mesh data with .obj suffix and drive the mesh using the style in some videos or something?

Have some problems when executing pip install -r requirements.txt. My current Python version is Python 3.11.9. Do I need to switch to an earlier version or install the latest libs not the version specified in requirements.txt or there's another approach I can use to resolve that?

can you provide the Pre-trained models?

Line 167 in model/trainer: where does Confidence come from?

Hi @chahuja,
Thanks for your impressive work on this, I really appreciate it.
While when I want to have a try of your codebase, it occurs the same problem on both training or inference by your pre-trained checkpoints like follows:

Actually, any case which involves the speaker "Oliver" would fail when training or inference. So could you please show me some suggestions on this?

I am looking forward to your reply.

Many thanks~

can you give some advice?

I followed the steps in the "Set Up Environment" section and a folder named pycasper was created. The problem is that this folder is outside "mix-stage" folder. So, files can't find the module. Should I transfer the pycasper directory inside "mix-stage" directory? Or should I transfer it to the model directory inside src directory inside "mix-stage"?

I downloaded the dataset for "cordon" and "lec_cosmic" only. I also downloaded the pretrained model and used the model:

exp_3659_cpk_JointLateClusterSoftStyle4_G_speaker_['corden',\ 'lec_cosmic']_model_JointLateClusterSoftStyle4_G_note_s2g_gst_mixgan15_weights.p

However, after the first model loads, it gives me this error:

/content/mix-stage/src/pycasper/BookKeeper.py:224: UserWarning: Counld not find results file
warnings.warn('Counld not find results file')
model.trainer.TrainerStyleClassifier selected
0% 0/10 [00:00<?, ?it/s]/content/drive/MyDrive/Gesture_Generation_Dataset/pats/data/processed/shelly/0.h5 pose/normalize
0% 0/10 [00:00<?, ?it/s]

Why does he need the speaker "shelly"?

This is the output:

When I run render.py using one of the pretrained models, I get this output:
Results Loaded
100% 10/10 [00:05<00:00, 1.77it/s]
100% 10/10 [00:10<00:00, 1.03s/it]
100% 10/10 [00:04<00:00, 2.22it/s]

Does this mean that the program finished correctly? If yes, then what and where is the output?