{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Change audio by detecting onset \n",
    "This notebook contains a method that could change the target video sound with a given audio."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "import IPython\n",
    "import os\n",
    "import numpy as np\n",
    "from moviepy.editor import *\n",
    "import librosa\n",
    "from IPython.display import Audio\n",
    "from IPython.display import Video"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read videos\n",
    "origin_video_path = 'demo-data/original.mp4'\n",
    "# conditional_video_path = 'demo-data/conditional.mp4'\n",
    "conditional_video_path = 'demo-data/dog_bark.mp4'\n",
    "\n",
    "ori_videoclip = VideoFileClip(origin_video_path)\n",
    "con_videoclip = VideoFileClip(conditional_video_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<video src=\"demo-data/original.mp4\" controls  width=\"640\" >\n",
       "      Your browser does not support the <code>video</code> element.\n",
       "    </video>"
      ],
      "text/plain": [
       "<IPython.core.display.Video object>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Video(origin_video_path, width=640)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<video src=\"demo-data/dog_bark.mp4\" controls  width=\"640\" >\n",
       "      Your browser does not support the <code>video</code> element.\n",
       "    </video>"
      ],
      "text/plain": [
       "<IPython.core.display.Video object>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Video(conditional_video_path, width=640)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get the audio track from video\n",
    "ori_audioclip = ori_videoclip.audio\n",
    "ori_audio, ori_sr = ori_audioclip.to_soundarray(), ori_audioclip.fps\n",
    "con_audioclip = con_videoclip.audio\n",
    "con_audio, con_sr = con_audioclip.to_soundarray(), con_audioclip.fps\n",
    "\n",
    "ori_audio = ori_audio.mean(-1)\n",
    "con_audio = con_audio.mean(-1)\n",
    "\n",
    "target_sr = 22050\n",
    "ori_audio = librosa.resample(ori_audio, orig_sr=ori_sr, target_sr=target_sr)\n",
    "con_audio = librosa.resample(con_audio, orig_sr=con_sr, target_sr=target_sr)\n",
    "\n",
    "ori_sr, con_sr = target_sr, target_sr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def detect_onset_of_audio(audio, sample_rate):\n",
    "    onsets = librosa.onset.onset_detect(\n",
    "        y=audio, sr=sample_rate, units='samples', delta=0.3)\n",
    "    return onsets\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "onsets = detect_onset_of_audio(ori_audio, ori_sr)\n",
    "plt.figure(dpi=100)\n",
    "\n",
    "time = np.arange(ori_audio.shape[0])\n",
    "plt.plot(time, ori_audio)\n",
    "plt.vlines(onsets, 0, ymax=0.8, colors='r')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Method\n",
    "The baseline is quite simple, and it has several steps:\n",
    "- Take the original video, and apply self-trained video onset detection model to detect the onset\n",
    "- Detect onsets of conditional waveform (encoded and decoded by our codebook) and clip single onset event from them as sound candicates\n",
    "- For each onset of original waveform, replace with conditional onset event randomly and then generate sound"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: CUDA_VISIBLE_DEVICES=9\n",
      "=> loading checkpoint 'checkpoints/EXP1/checkpoint_ep70.pth.tar'\n",
      "=> loaded checkpoint 'checkpoints/EXP1/checkpoint_ep70.pth.tar' (epoch 70)\n"
     ]
    }
   ],
   "source": [
    "%env CUDA_VISIBLE_DEVICES=9\n",
    "import argparse\n",
    "import numpy as np\n",
    "import os\n",
    "import sys\n",
    "import time\n",
    "from tqdm import tqdm\n",
    "from collections import OrderedDict\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader\n",
    "from torch.utils.tensorboard import SummaryWriter\n",
    "\n",
    "\n",
    "from config import init_args\n",
    "import data\n",
    "import models\n",
    "from models import *\n",
    "from utils import utils, torch_utils\n",
    "\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "\n",
    "net = models.VideoOnsetNet(pretrained=False).to(device)\n",
    "resume = 'checkpoints/EXP1/checkpoint_ep70.pth.tar'\n",
    "net, _ = torch_utils.load_model(resume, net, device=device, strict=True)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torchvision.transforms as transforms\n",
    "from PIL import Image\n",
    "\n",
    "\n",
    "vision_transform_list = [\n",
    "    transforms.Resize((128, 128)),\n",
    "    transforms.CenterCrop((112, 112)),\n",
    "    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])\n",
    "]\n",
    "video_transform = transforms.Compose(vision_transform_list)\n",
    "\n",
    "def read_image(frame_list):\n",
    "    imgs = []\n",
    "    convert_tensor = transforms.ToTensor()\n",
    "    for img_path in frame_list:\n",
    "        image = Image.open(img_path).convert('RGB')\n",
    "        image = convert_tensor(image)\n",
    "        imgs.append(image.unsqueeze(0))\n",
    "    # (T, C, H ,W)\n",
    "    imgs = torch.cat(imgs, dim=0).squeeze()\n",
    "    imgs = video_transform(imgs)\n",
    "    imgs = imgs.permute(1, 0, 2, 3)\n",
    "    # (C, T, H ,W)\n",
    "    return imgs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "# process videos into frames and read them\n",
    "import glob\n",
    "\n",
    "save_path = 'demo-data/original_frames'\n",
    "if os.path.exists(save_path):\n",
    "    os.system(f'rm -rf {save_path}')\n",
    "os.makedirs(save_path)\n",
    "command = f'ffmpeg -v quiet -y -i \\\"{origin_video_path}\\\" -f image2 -vf \\\"scale=-1:360,fps=15\\\" -qscale:v 3 \\\"{save_path}\\\"/frame%06d.jpg'\n",
    "os.system(command)\n",
    "\n",
    "frame_list = glob.glob(f'{save_path}/*.jpg')\n",
    "frame_list.sort()\n",
    "frame_list = frame_list[:2 * 15]\n",
    "\n",
    "frames = read_image(frame_list)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "inputs = {\n",
    "    'frames': frames.unsqueeze(0).to(device)\n",
    "}\n",
    "pred = net(inputs).squeeze()\n",
    "pred = torch.sigmoid(pred).data.cpu().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "def postprocess_video_onsets(probs, thres=0.5, nearest=5):\n",
    "    # import pdb; pdb.set_trace()\n",
    "    video_onsets = []\n",
    "    pred = np.array(probs, copy=True)\n",
    "    while True:\n",
    "        max_ind = np.argmax(pred)\n",
    "        video_onsets.append(max_ind)\n",
    "        low = max(max_ind - nearest, 0)\n",
    "        high = min(max_ind + nearest, pred.shape[0])\n",
    "        pred[low: high] = 0\n",
    "        if (pred > thres).sum() == 0:\n",
    "            break\n",
    "    video_onsets.sort()\n",
    "    video_onsets = np.array(video_onsets)\n",
    "    return video_onsets\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "# video_onsets = (np.nonzero(pred > 0.5)[0] / 15 * ori_sr).astype(int)\n",
    "video_onsets = postprocess_video_onsets(pred, thres=0.5, nearest=4)\n",
    "video_onsets = (video_onsets / 15 * ori_sr).astype(int)\n",
    "plt.figure(dpi=100)\n",
    "\n",
    "time = np.arange(ori_audio.shape[0])\n",
    "plt.plot(time, ori_audio)\n",
    "plt.vlines(video_onsets, 0, ymax=0.8, colors='r')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([-0.06068027, -0.0599093 , -0.05623583, -0.01206349])"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(onsets - video_onsets) / ori_sr\n",
    "# video_onsets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_onset_audio_range(audio_len, onsets, i):\n",
    "    if i == 0:\n",
    "        prev_offset = int(onsets[i] // 3)\n",
    "    else:\n",
    "        prev_offset = int((onsets[i] - onsets[i - 1]) // 3)\n",
    "\n",
    "    if i == onsets.shape[0] - 1:\n",
    "        post_offset = int((audio_len - onsets[i]) // 4 * 2)\n",
    "    else:\n",
    "        post_offset = int((onsets[i + 1] - onsets[i]) // 4 * 2)\n",
    "    return prev_offset, post_offset\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "# ori_onsets = detect_onset_of_audio(ori_audio, ori_sr)\n",
    "con_onsets = detect_onset_of_audio(con_audio, con_sr)\n",
    "\n",
    "np.random.seed(2022)\n",
    "gen_audio = np.zeros_like(ori_audio)\n",
    "for i in range(video_onsets.shape[0]):\n",
    "    prev_offset, post_offset = get_onset_audio_range(int(con_sr * 2), video_onsets, i)\n",
    "    j = np.random.choice(con_onsets.shape[0])\n",
    "    prev_offset_con, post_offset_con = get_onset_audio_range(con_audio.shape[0], con_onsets, j)\n",
    "    prev_offset = min(prev_offset, prev_offset_con)\n",
    "    post_offset = min(post_offset, post_offset_con)\n",
    "    gen_audio[video_onsets[i] - prev_offset: video_onsets[i] + post_offset] = con_audio[con_onsets[j] - prev_offset: con_onsets[j] + post_offset]\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x400 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "plt.figure(dpi=100)\n",
    "time = np.arange(gen_audio.shape[0])\n",
    "plt.plot(time, gen_audio)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save audio\n",
    "import soundfile as sf\n",
    "sf.write('data/gen_audio.wav', gen_audio, ori_sr)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "t:  58%|█████▊    | 26/45 [00:41<00:05,  3.45it/s, now=None]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Moviepy - Building video data/generate.mp4.\n",
      "MoviePy - Writing audio in generateTEMP_MPY_wvf_snd.mp3\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "t:  58%|█████▊    | 26/45 [00:42<00:05,  3.45it/s, now=None]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MoviePy - Done.\n",
      "Moviepy - Writing video data/generate.mp4\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "t:  58%|█████▊    | 26/45 [01:03<00:05,  3.45it/s, now=None]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Moviepy - Done !\n",
      "Moviepy - video ready data/generate.mp4\n"
     ]
    }
   ],
   "source": [
    "gen_audioclip = AudioFileClip(\"data/gen_audio.wav\")\n",
    "gen_videoclip = ori_videoclip.set_audio(gen_audioclip)\n",
    "gen_videoclip.write_videofile('data/generate.mp4')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<video src=\"data/generate.mp4\" controls  width=\"640\" >\n",
       "      Your browser does not support the <code>video</code> element.\n",
       "    </video>"
      ],
      "text/plain": [
       "<IPython.core.display.Video object>"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Video('data/generate.mp4', width=640)\n"
   ]
  }
 ],
 "metadata": {
  "interpreter": {
   "hash": "419ed25a44e8f5205333d07bc5a26d3abb4bd07afa4dac02924f75b129c3e2d9"
  },
  "kernelspec": {
   "display_name": "Python 3.8.8 ('AVanalogy')",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}