Model_Avatar.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Model_Avatar.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/dream80/DeepFaceLab_Colab/blob/master/Model_Avatar.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KWQMLYlFrTXz",
        "colab_type": "text"
      },
      "source": [
        "# 简介\n",
        "\n",
        "无需多说，一路运行即可！\n",
        "\n",
        "此脚本由**[托尼是塔克](https://wxp123.me)**创建，仅用于学习，请勿滥用。如有疑问，拉到最后！"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AmGMbWZef1XI",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 查看设备，是K80还是T4,如果是K80...！\n",
        "! /opt/bin/nvidia-smi"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tpey8HLsPVpP",
        "colab_type": "text"
      },
      "source": [
        "**注意：AVATAR模型的data_src.mp4视频需要1:1,方形视频**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ldpC3PjRrYYh",
        "colab_type": "text"
      },
      "source": [
        "# 第一步 准备好workspace\n",
        "这一步你可以选着两种方式\n",
        "1. 使用默认的workspace，你无需自己上传，仅用于熟悉操作。\n",
        "2. 通过Google Drive （谷歌云盘）上传自己的workspace到指定目录。\n",
        "\n",
        "\n",
        "\n",
        "谷歌云盘地址：https://drive.google.com/drive/my-drive  \n",
        "注意：谷歌网址现在都需要“科学上网”才能访问。  "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D7toZxhT4J9W",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#挂载谷歌云盘\n",
        "#点击链接授权，复制授权码，填入框框，然后回车。\n",
        "\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive', force_remount=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rDUpbJYR54dw",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#创建DeepFaceLab目录，并且进入目录\n",
        "%cd /content/drive/My Drive/\n",
        "!mkdir DeepFaceLab\n",
        "%cd /content/drive/My Drive/DeepFaceLab/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cNsq0I98KvHl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 下面的1和2 不用同时点，二选一！！！！\n",
        "# 1. 练手可以使用这一行，直接git clone一个workspace\n",
        "!git clone https://github.com/dream80/DFLWorkspace.git workspace"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w6ViQYB_Cupc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 2.正常情况肯定是通过谷歌云盘上传自己的workspace，上传前压缩一下，然后用下面的命令解压。\n",
        "# 注意一定要上传到DeepFaceLab目录下。\n",
        "!unzip  workspace.zip\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yA4nziLPtAf9",
        "colab_type": "text"
      },
      "source": [
        "#第二步 安装DeepFaceLab\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4CbWbLzHzqTQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 获取DFL源代码v1.6.1稳定版\n",
        "# 注意要获取新的版本，记得通过云盘把DeepFaceLab_Colab目录删除，然后执行以下命令\n",
        "\n",
        "!git clone -b v1.6.2 https://github.com/dream80/DeepFaceLab_Colab.git\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5MzbdSPEzsF6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 进入DeepFaceLab_Colab目录\n",
        "%cd /content/drive/My Drive/DeepFaceLab/DeepFaceLab_Colab\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yd11UhiUvUfP",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "# 安装Python依赖\n",
        "!pip install -r requirements_colab.txt\n",
        "!pip install --upgrade scikit-image"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9G9s5gJrty-x",
        "colab_type": "text"
      },
      "source": [
        "# 第三步. 提取脸部"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "97VTmSmkEk2j",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "# 确保路径正确，进入DeepFaceLab_Colab\n",
        "\n",
        "%cd /content/drive/My Drive/DeepFaceLab/DeepFaceLab_Colab"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "w8qxq7jAxGrQ",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Src 视频转图片\n",
        "\n",
        "!python main.py videoed extract-video --input-file ../workspace/data_src.mp4 --output-dir ../workspace/data_src/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hpTukPbSxbGP",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Src 提取脸部图片\n",
        "\n",
        "!python main.py extract --input-dir ../workspace/data_src --output-dir ../workspace/data_src/aligned --detector s3fd --face-type mark_only"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kWQsUODU8Ftg",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Src排序，可以通过谷歌云盘查看结果，删除不良图片\n",
        "\n",
        "!python main.py sort --input-dir ../workspace/data_src/aligned --by hist"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "z35vHpqaw8t-",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Dst视频转图片\n",
        "\n",
        "!python main.py videoed extract-video --input-file ../workspace/data_dst.mp4 --output-dir ../workspace/data_dst/"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EbAY9bJvxDBK",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Dst提取脸部图片\n",
        "\n",
        "!python main.py extract --input-dir ../workspace/data_dst --output-dir ../workspace/data_dst/aligned --detector s3fd --face-type full_face_no_align"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PJs_yo-F8IBS",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#Dst排序，可以通过谷歌云盘查看结果，删除不良图片\n",
        "\n",
        "!python main.py sort --input-dir ../workspace/data_dst/aligned --by hist"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QxO1kNO_uRTF",
        "colab_type": "text"
      },
      "source": [
        "# 第四步. 训练模型\n",
        "\n",
        "\n",
        "*  支持H128,SAE,DF, LIAEF128，AVATAR等模型，根据自己的情况选择模型。\n",
        "*  训练开始是需要配置参数，记得开启预览，其他参数根据自己情况选择，使用默认参数的话直接回车即可。\n",
        "*  不想训练了可以点击停止，停止时会抛出异常，但是没什么关系。下次可以继续训练"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7n3ZpkQd-dZj",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 确保路径正确，进入DeepFaceLab_Colab\n",
        "\n",
        "%cd /content/drive/My Drive/DeepFaceLab/DeepFaceLab_Colab"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NvP37RbjsN8V",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "3fe6043e-ae94-4e4e-fc42-3270af9b5a40"
      },
      "source": [
        "# 开始训练 AVATAR 模型 \n",
        "\n",
        "!python main.py train --training-data-src-dir ../workspace/data_src/aligned --training-data-dst-dir ../workspace/data_dst/aligned --model-dir ../workspace/model --model AVATAR --no-preview"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "^C\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oVfZpBR00gBG",
        "colab_type": "text"
      },
      "source": [
        "**模型预览方法请参考：**\n",
        "https://www.wxp123.me/p/329\n",
        "\n",
        "简单描述：左侧->文件->/driver/..../model/ *preview*.jpg->双击  \n",
        "  \n",
        "这种方法比在脚本中预览和在网盘预览都要方便，并且无需启动预览配置！"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m2Ba96cVuS0K",
        "colab_type": "text"
      },
      "source": [
        "# 第五步. 转换输出\n",
        "\n",
        "默认使用SAE进行转换，如果需要其他模型，请修改命令中的参数。\n",
        "\n",
        "比如将 --model SAE 修改为 --model H128"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "__aUSMa80-_M",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#用src中的脸替换dst的脸 \n",
        "!python main.py convert --input-dir ../workspace/data_dst --output-dir ../workspace/data_dst/merged --aligned-dir ../workspace/data_dst/aligned --model-dir ../workspace/model --model AVATAR"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "symb8EIc1Csf",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "#把替换好的图片转换成视频\n",
        "\n",
        "!python main.py videoed video-from-sequence --input-dir ../workspace/data_dst/merged --output-file ../workspace/result.mp4 --reference-file ../workspace/data_dst.mp4"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cVfp8xbY5ykL",
        "colab_type": "text"
      },
      "source": [
        "#第六步. 继续训练\n",
        "当你第二次开始训练，或者掉线之后继续训练时并不需要执行上面所有的步骤。只需要下面简单的几个步骤。\n",
        "1. 挂载云盘\n",
        "2. 安装依赖\n",
        "3. 开始训练"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "veL_8afb6WcP",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#挂载谷歌云盘\n",
        "#点击链接授权，复制授权码，填入框框，然后回车。\n",
        "\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive', force_remount=True)\n",
        "\n",
        "#重新覆盖解压，提升加载速度，可节省很多时间\n",
        "%cd /content/drive/My Drive/DeepFaceLab/\n",
        "!unzip  -o workspace.zip\n",
        "\n",
        "\n",
        "# 进入DeepFaceLab_Colab目录\n",
        "%cd /content/drive/My Drive/DeepFaceLab/DeepFaceLab_Colab\n",
        "\n",
        "# 安装Python依赖\n",
        "!pip install -r requirements_colab.txt\n",
        "!pip install --upgrade scikit-image\n",
        "\n",
        "# 开始训练SAE ，如果是其他模型，修改后面的参数即可，比如H128。\n",
        "\n",
        "!python main.py train --training-data-src-dir ../workspace/data_src/aligned --training-data-dst-dir ../workspace/data_dst/aligned --model-dir ../workspace/model --model AVATAR --no-preview"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "siaPGJ1QVacj",
        "colab_type": "text"
      },
      "source": [
        "# 工具\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WGLTllAHYAPJ",
        "colab_type": "text"
      },
      "source": [
        "使用脚本练模型的时候常常会需要上传，解压，压缩，下载等，下面为大家提供一个解压和压缩的脚本。如果需要对其他文件操作，改一下路径，改一下文件名就可以了。"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e2Y0j4dBVaG_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 解压workspace.zip，假设这个文件在DeepFaceLab目录下。\n",
        "\n",
        "%cd /content/drive/My Drive/DeepFaceLab/\n",
        "\n",
        "#常规解压，如果之前有文件会询问是否覆盖，输入A表示覆盖\n",
        "!unzip workspace.zip\n",
        "\n",
        "#覆盖解压，直接覆盖，不需要询问。\n",
        "#unzip -o workspace.zip \n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n4jduhJMWXia",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# 压缩worksapce ，把压缩包保存在DeepFaceLab目录下\n",
        "%cd /content/drive/My Drive/DeepFaceLab/\n",
        "!zip -r workspace.zip  workspace"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5H85SsIW1PI3",
        "colab_type": "text"
      },
      "source": [
        "# 其他\n",
        "\n",
        " 谷歌云地址：https://drive.google.com/drive/my-drive\n",
        "\n",
        " 作者邮箱 ：wpgdream@gmail.com\n",
        " \n",
        " 使用教程：https://www.wxp123.me\n",
        " \n",
        " \n",
        "不求别的，只求在github给个小星星^_^\n",
        "\n",
        "https://github.com/dream80/DeepFaceLab_Colab\n",
        "\n",
        "右上角star 谢谢!!!"
      ]
    }
  ]
}