Release Version 1.0 ByteDance Piano Transcription Colab

Project Los Angeles
Tegridy Code 2023

tegridy-tools/notebooks/ByteDance_Piano_Transcription.ipynb

@@ -0,0 +1,115 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "private_outputs": true,
+      "provenance": [],
+      "gpuType": "T4"
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU"
+  },
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "source": [
+        "***\n",
+        "\n",
+        "# ByteDance Piano Transcription (Ver 1.0)\n",
+        "\n",
+        "***\n",
+        "\n",
+        "## https://github.com/bytedance/piano_transcription\n",
+        "\n",
+        "## https://github.com/azuwis/pianotrans\n",
+        "\n",
+        "***\n",
+        "\n",
+        "### Project Los Angeles\n",
+        "### Tegridy Code 2023\n",
+        "\n",
+        "***"
+      ],
+      "metadata": {
+        "id": "X7JFKeW7HTr4"
+      }
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# (SETUP ENVIRONMENT)"
+      ],
+      "metadata": {
+        "id": "MFc-x5DvH_4k"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# @title Install dependencies\n",
+        "!pip install librosa==0.8.1\n",
+        "!pip install piano_transcription_inference"
+      ],
+      "metadata": {
+        "id": "cirk7Velwb6W",
+        "cellView": "form"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# (TRANSCRIBE)\n",
+        "\n",
+        "## 1) The first run may take a long time due to the model being downloaded from Zenodo which is quite slow usually\n",
+        "\n",
+        "## 2) Input audio must be in WAV format"
+      ],
+      "metadata": {
+        "id": "4wd0mKVoH1dF"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# @title Transcribe\n",
+        "full_path_to_input_WAV = \"/content/input.wav\" # @param {type:\"string\"}\n",
+        "full_path_to_output_MIDI = \"/content/output.mid\" # @param {type:\"string\"}\n",
+        "\n",
+        "from piano_transcription_inference import PianoTranscription, sample_rate, load_audio\n",
+        "\n",
+        "# Load audio\n",
+        "(audio, _) = load_audio(full_path_to_input_WAV, sr=sample_rate, mono=True)\n",
+        "\n",
+        "# Transcriptor\n",
+        "transcriptor = PianoTranscription(device='cuda')    # 'cuda' | 'cpu'\n",
+        "\n",
+        "# Transcribe and write out to MIDI file\n",
+        "transcribed_dict = transcriptor.transcribe(audio, full_path_to_output_MIDI)"
+      ],
+      "metadata": {
+        "id": "laaR9o9nwDzt",
+        "cellView": "form"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# Congrats! You did it! :)"
+      ],
+      "metadata": {
+        "id": "0jKlAwVHH4G4"
+      }
+    }
+  ]
+}