sylyt62
sylyt62
Gotcha. Thx~
Does it support the newest 3.1.2?
Oh I see, you select 105 faces out of 9976 faces in total, and calculate the barycenter of each face as the landmark? Why are you doing that instead of...
I tried this repo. I used the official syncnet_v2.model, but the performance is really bad on my mandarin dataset. That leads to my question above lol...
Hey @ChengsongLu , I got similar results as yours.. I trained it with CMLR dataset on a 4090 for 7 days. 
 Same here... Trained on 4090 for 7 days. @zhangziliang04 Any tips to get a further drop? My wechat: tybs8k
Actually nothing changes when changing this parameter. The rotation keeps the same
I mean, after using the default flat_hand_mean=False, the displayed pose keeps almost the same as above pictures.
I hardcode the `num_works` of `train_loader` to 0 and it works.
Hey! Have you resolved this issue? I got worse rendering results than you.. Can't even tell where the bone is. But the validation images are correct  