fastai2_audio_xresnet18.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.6"
    },
    "colab": {
      "name": "Coo_Fastai2_Audio.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/AdPostma/open_collaboration_on_audio_classification/blob/master/Coo_Fastai2_Audio.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eVUyVIVHjOiN",
        "colab_type": "text"
      },
      "source": [
        "##Maraques Coo calls model with Fastai2_audio \n",
        "\n",
        "Adapted from Fastai2_audio: 02_tutorial. Thanks to rbracco"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Pjq-SLd6pqKL",
        "colab_type": "text"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6zBubzcFgdLe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install git+https://github.com/fastai/fastai2\n",
        "!pip install git+https://github.com/fastai/fastcore\n",
        "!pip install packaging\n",
        "!pip install git+https://github.com/rbracco/fastai2_audio.git"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sVg8j4sud4g1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "#hide\n",
        "%reload_ext autoreload\n",
        "%autoreload 2\n",
        "%matplotlib inline"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j4LXndMId5Nc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from fastai2.torch_basics import *\n",
        "from fastai2.basics import *\n",
        "from fastai2.data.all import *\n",
        "from fastai2.callback.all import *\n",
        "from fastai2.vision.all import *\n",
        "\n",
        "from fastai2_audio.core import *\n",
        "from fastai2_audio.augment import *\n",
        "\n",
        "import torchaudio"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "REqKRkcBkSnJ",
        "colab_type": "text"
      },
      "source": [
        "### Loading maraques coo soundfiles 24.4 kHz version"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L0v40WLOgPC9",
        "colab_type": "code",
        "outputId": "97c77514-ec69-411b-8306-58eb59952e2e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        }
      },
      "source": [
        "Coo = untar_data('https://storage.googleapis.com/ml-animal-sounds-datasets/macaques_24414Hz.zip')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eCKuekLreXW_",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "x = AudioGetter(\"\", recurse=True, folders=None)\n",
        "Coo_files  = x(Coo)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XHyq0sn3ks2n",
        "colab_type": "text"
      },
      "source": [
        "### Creating configuration file and item transformations"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "7ClI06uyfhTD",
        "colab": {}
      },
      "source": [
        "#crop 750ms from the signal and turn it to a MelSpectrogram with no augmentation\n",
        "cfg_voice = AudioConfig.Voice(sample_rate=24414,n_fft=1024, hop_length=140,f_max=12207)\n",
        "a2s = AudioToSpec.from_cfg(cfg_voice)\n",
        "crop_750ms = CropSignal(750)\n",
        "tfms = [crop_750ms, a2s]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oJJY4ut-lbdr",
        "colab_type": "text"
      },
      "source": [
        "### Show example of Mellspectrogram\n",
        "(show_batch() is broken in Fastai-2 Audio)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "YCrCY-NIfiVQ",
        "outputId": "4ff8555c-9141-4762-b23c-99453a6e9bd4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        }
      },
      "source": [
        "pipe = Pipeline([AudioTensor.create, crop_750ms,a2s])\n",
        "pipe(Coo_files[1]).show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1h7Ff8hal3j0",
        "colab_type": "text"
      },
      "source": [
        "### Create audio dataloaders"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KUDSYGdGgZya",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "auds = DataBlock(blocks=(AudioBlock, CategoryBlock),  \n",
        "                 get_items=get_audio_files, \n",
        "                 splitter=RandomSplitter(seed=42),\n",
        "                 item_tfms = tfms,\n",
        "                 get_y=lambda x: str(x).split('/')[-1][:2])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qLAX7mKlgZyc",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "dbunch = auds.dataloaders(Coo, bs=64)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IZopK-xwgZye",
        "colab_type": "code",
        "outputId": "e9b5f7a3-37b5-4466-aff8-7c8f61ac62e1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "dbunch.one_batch()[0].shape"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([64, 1, 128, 131])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hjyXGBQsmFlP",
        "colab_type": "text"
      },
      "source": [
        "### Create and train learner"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J43YnCyfgZyh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# credit to Kevin Bird and Hiromi Suenaga for these two lines to adjust a CNN model to take 1 channel input\n",
        "def alter_learner(learn, channels=1):\n",
        "    learn.model[0][0].in_channels=channels\n",
        "    learn.model[0][0].weight = torch.nn.parameter.Parameter(learn.model[0][0].weight[:,1,:,:].unsqueeze(1))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "91eHXUSdgZyk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "learn = Learner(dbunch, \n",
        "                xresnet18(),\n",
        "                torch.nn.CrossEntropyLoss(), \n",
        "                metrics=[error_rate])\n",
        "nchannels = dbunch.one_batch()[0].shape[1]\n",
        "alter_learner(learn, nchannels)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KUOKlZtSgFaX",
        "colab_type": "code",
        "outputId": "fc6fd051-f8a6-4692-832e-259662e3f16b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 267
        }
      },
      "source": [
        "learn.fine_tune(5)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>error_rate</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.537560</td>\n",
              "      <td>0.407273</td>\n",
              "      <td>0.133150</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>error_rate</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>0.120560</td>\n",
              "      <td>0.192001</td>\n",
              "      <td>0.061084</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.079518</td>\n",
              "      <td>0.270885</td>\n",
              "      <td>0.096774</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.034710</td>\n",
              "      <td>0.018015</td>\n",
              "      <td>0.005491</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.016162</td>\n",
              "      <td>0.012601</td>\n",
              "      <td>0.004118</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.007578</td>\n",
              "      <td>0.010043</td>\n",
              "      <td>0.002745</td>\n",
              "      <td>00:22</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    }
  ]
}