搬运一个WF288 LIAE-UDT的高参底丹，可用于实时直播模型训练
==---------------------- Model Options ----------------------==== ==
== resolution: 288 ==
== face_type: wf ==
== models_opt_on_gpu: True ==
== archi: liae-udt ==
== ae_dims: 384 ==
== e_dims: 92 ==
== d_dims: 72 ==
== d_mask_dims: 24 ==
== masked_training: True ==
== eyes_mouth_prio: True ==
== uniform_yaw: False ==
== blur_out_mask: False ==
== adabelief: True ==
== lr_dropout: n ==
== random_warp: True ==
== random_hsv_power: 0.0 ==
== true_face_power: 0.0 ==
== face_style_power: 0.0 ==
== bg_style_power: 0.0 ==
== ct_mode: none ==
== clipgrad: False ==
== pretrain: False ==
== autobackup_hour: 12 ==
== write_preview_history: False ==
== target_iter: 2500000 ==
== random_src_flip: False ==
== random_dst_flip: False ==
== batch_size: 14 ==
== gan_power: 0.0 ==
== gan_patch_size: 36 ==
== gan_dims: 16 ==
== ==
模型参数如上，可用于直播换脸模型底丹
主播模型训练方法简单说明：整理好你的src素材，找一个dst数据集，然后解压模型，直接运行训练，修改batch_size这个参数，根据自己显卡修改
然后训练10万迭代左右，如果觉得src不够像，可以删除inter_AB.npy这个文件，继续训练
然后关闭随机扭曲，开gan收尾
下载地址：
**** 本内容需购买 ****
页:
[1]