BugNet

推理结果为噪声

3

> solo_solo_pose1_512x512_3_1438_noaudio.mp4 > 我们复现了您反馈的问题。 触发条件没有正确加载我们预训练权重，导致模型参数跟我们的任务不匹配。 您可以重新下载我们的预训练模型，并确保其位于正确路径下，详情参考README.md。 https://huggingface.co/ZJYang/AniPortrait/tree/main 感谢您的回复，请问复现的这段视频是哪部分模型参数没有加载正确呢？我参考了本repo中的预训练模型目录，我的目录如下，看上去和您给的readme一致：  _Originally posted by @fredkingdom in https://github.com/Zejun-Yang/AniPortrait/issues/35#issuecomment-2026791762_

I've notice that `vico_challenge_baseline/PIRender/inference.py` outputs the ground truth video rather than the generated one. It should be a problem with `write2video()` function. After I replace this function with the original...

I've notice there seems to be no modifications to `listenr_generator.py` in vico baseline method. Does it mean that this method has the same listener training pipeline as the original baseline?...

Bug in train_stage_1.py, causing list out of range

2

In train_stage_1.py, line 167, the correct code should be `sample_idx = [random.randint(0, dataset_len-1) for _ in range(2)]` instead of `sample_idx = [random.randint(0, dataset_len) for _ in range(2)]`, since if you...

Thanks for open-sourcing such a wonderful work! I notice that in Fig.1 of the technique report, the neutral mesh is used to calculated the mesh offset during audio2lmk training. It...

Hi, thanks for the open source! I have several questions regarding dataset processing, model training and evaluation. 1. It seems that EMOCA feature and HuBERT features are required in `python...