通过官方数据集训练不同脸型、角度、光影、表情等训练了470w迭代,主要针对直播使用
AMP在SAEHD的基础上删减和引入了新的参数,使其更加适合于实时换脸的场景。如果你需要能灵活设置参数,实现高质量的视频换脸,那么肯定用SAEHD。
使用方法:所有文件放到model文件夹,删除模型此文件夹中的_AMP_inter_dst.npy文件写入遮罩,将自己src、dst放到aligned文件夹,选择train AMP进行训练
loss跑到0.2左右就比较清晰了,追求完美的继续炼
练好了模型导出选择6) export AMP as dfm
把dfm文件放到live的dfm_models文件夹下就可以用了
注意:dst文件尽量放被替换人物的图片,这样模型训练出来后看起来更像
如果你训练模型是为了应用在DeepFaceLive中实现实时换脸,推荐AMP,因为带变形因子。
================== Model Summary =================== == == == Model name: AMPSRCSRC_AMP == == == == Current iteration: 4702331 == == == ==---------------- Model Options -----------------== == == == resolution: 224 == == face_type: wf == == models_opt_on_gpu: True == == ae_dims: 512 == == inter_dims: 1024 == == e_dims: 64 == == d_dims: 64 == == d_mask_dims: 22 == == morph_factor: 0.5 == == uniform_yaw: False == == blur_out_mask: False == == lr_dropout: n == == random_warp: True == == ct_mode: none == == clipgrad: False == == autobackup_hour: 0 == == write_preview_history: False == == target_iter: 0 == == random_src_flip: False == == random_dst_flip: False == == batch_size: 8 == == gan_power: 0.0 == == gan_patch_size: 28 == == gan_dims: 16 == == == ==------------------ Running On ------------------==
|