|
|
一位在高墙之内，身穿囚服的罪犯，戴着手铐脚镣接受警察的审问。他如实的交代了他的罪行，但其实他根本没有交代过，是用AI软件合成了这段视频...
大多数高质量的deepfakes都是用DeepFaceLab制作的。DFL为创建deepfakes提供了端到端的解决方案，从数据收集和管理到模型训练和最终视频输出。
深度伪造，简称深伪，是英文“deep learning”和“fake”的混成词，专指基于人工智能的人体图像合成技术的应用。此技术可将已有的图像或影片叠加至目标图像或影片上。伪造面部表情并将其呈现至目标影片的技术在2016年出现，此技术允许近实时地伪造文书现有2D视频中的面部表情。 维基百科
DeepFaceLab NVIDIA RTX 3000 系列构建 支持（并需要）NVIDIA RTX 3000 系列 GPU 或更高版本（即 4000、5000、6000 等）
DeepFaceLab英文原版下载地址
链接：https://pan.baidu.com/s/1DBnKY53ubZol6VkWehuRIw
提取码：bmkj
启动硬件GPU加速
虚拟内存设置
加载读取人脸数据过程中，需要占用大量内存（此处指内存，不是显存），很可能内存耗尽导致程序停止，报出“页面文件太小，无法完成操作”"memory error"等错误。
需要设置较大虚拟内存，用一部分硬盘空间来临时当做内存使用。
设置方法百度“如何设置虚拟内存”
6) 训练AMP模型 train AMP
开始训练
================== Model Summary ===================
== ==
== Model name: liu256-256_AMP ==
== ==
== Current iteration: 627873 ==
== ==
==---------------- Model Options -----------------==
== ==
== resolution: 256 ==
== face_type: wf ==
== models_opt_on_gpu: True ==
== ae_dims: 256 ==
== inter_dims: 64 ==
== 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: False ==
== ct_mode: none ==
== clipgrad: False ==
== autobackup_hour: 0 ==
== write_preview_history: False ==
== target_iter: 1000000 ==
== random_src_flip: False ==
== random_dst_flip: False ==
== batch_size: 8 ==
== gan_power: 0.0 ==
== gan_patch_size: 28 ==
== gan_dims: 16 ==
== ==
==------------------ Running On ------------------==
== ==
== Device index: 0 ==
== Name: NVIDIA GeForce RTX 3070 ==
== VRAM: 5.33GB ==
== ==
====================================================
参考：
https://www.aibl.vip/thread-499-1-1.html
https://zhuanlan.zhihu.com/p/648261832
deepfacelab AI换脸训练参数详解
DeepFaceLab软件开发作者对于各项参数的官方介绍 (出处: deepfacelab中文网)
Deepfacelab模型训练，人人都可以训练一个专属自己的模型
DeepFaceLab 2.0 指南 (出处: ZhaTV)
发现一个错误
Traceback (most recent call last):
File "G:\DeepFaceLab\DeepFaceLab_NV30_221111_CN\_internal\DeepFaceLab\core\joblib\SubprocessorBase.py", line 71, in _subprocess_run
result = self.process_data (data)
File "G:\DeepFaceLab\DeepFaceLab_NV30_221111_CN\_internal\DeepFaceLab\merger\InteractiveMergerSubprocessor.py", line 118, in process_data
raise Exception( f'Error while merging file [{filepath}]: {e_str}' )
Exception: Error while merging file [F:\DFL\DeepFaceLab_NV30_221111_CN\workspace\data_dst\00008.png]: Traceback (most recent call last):
File "G:\DeepFaceLab\DeepFaceLab_NV30_221111_CN\_internal\DeepFaceLab\merger\InteractiveMergerSubprocessor.py", line 112, in process_data
frame_info=frame_info)
File "G:\DeepFaceLab\DeepFaceLab_NV30_221111_CN\_internal\DeepFaceLab\merger\MergeMasked.py", line 325, in MergeMasked
img_bgr_uint8 = imagelib.normalize_channels (img_bgr_uint8, 3)
File "G:\DeepFaceLab\DeepFaceLab_NV30_221111_CN\_internal\DeepFaceLab\core\imagelib\common.py", line 15, in normalize_channels
img_shape_len = len(img.shape)
AttributeError: 'NoneType' object has no attribute 'shape'
将DeepFaceLab_NV30_221111_CN移动到[F:\DFL\下就好了。
|
|
