MagicDance
MagicDance copied to clipboard
Boese0601
[ICML 2024] MagicPose(also known as MagicDance): Realistic Human Poses and Facial Expressions Retargeting with Identity-aware Diffusion
Hi, thank you for this awesome project, just wanted to know. Why is the generated image quality low compared to the examples shown in the readme file? I am sharing...
Thanks for your Great Work. It's mentioned that you conduncted a expriment on self-attention, reference image and the noisy image are both forwarded through the diffusion UNet with their self-attention...
你好，感谢开源。 针对开源的模型权重model_state-110000，我理解是包含了外观模块、姿态控制模块和去噪Unet模块， 没有包含时序运动模块。 开源的权重中没有训练运动模块，是这样吗? 开源的代码中是否支持训练运动模块？网络结构是否包含了运动模块？
Greetings ! My attempt to run the code failed duo out of memory while loading the model. I am using an RTX 3060 with 12GB What are the minimum specs...
Could you please clarify whether the input to the appearance control model is the latent of the reference image or the noisy latent after DDIM inversion processing? thanks!!
I used this image and "001" pose to generate gif. However, Generated gif has different face ID with input image.   cmd: bash scripts/inference_any_image_pose.sh inference_any_image_pose.sh is...
Great job, but could we at least have a schematic diagram of the data? Downloading 33G of data is too slow, and isn't providing inference code image by image a...
Does anyone know how to load weight from control_v11p_sd15_openpose.pth ? It appears that train_tiktok.py doesn't have code to load weight from this pth file.
So, optimizer_state_latest.th is not available? as in Originally posted by @Boese0601 in https://github.com/Boese0601/MagicDance/issues/18#issuecomment-1975676943