{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"In this notebook, we'll use a pre-trained machine learning model to generate a submission to the [BirdClef2023 competition](https://www.kaggle.com/c/birdclef-2023).  The goal of the competition is to identify Eastern African bird species by sound.","metadata":{}},{"cell_type":"markdown","source":"## Step 1: Imports","metadata":{}},{"cell_type":"code","source":"import tensorflow as tf\nimport tensorflow_hub as hub\nimport tensorflow_io as tfio\n\nimport pandas as pd\nimport numpy as np\nimport librosa\nimport glob\n\nimport csv\nimport io\n\nfrom IPython.display import Audio","metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2023-06-08T09:04:29.432179Z","iopub.execute_input":"2023-06-08T09:04:29.432776Z","iopub.status.idle":"2023-06-08T09:04:41.354852Z","shell.execute_reply.started":"2023-06-08T09:04:29.432730Z","shell.execute_reply":"2023-06-08T09:04:41.353319Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Load a sample audio files from two different species\naudio_abe, sr_abe = librosa.load(\"/kaggle/input/birdclef-2023/train_audio/abethr1/XC128013.ogg\")\n","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:38:20.308748Z","iopub.execute_input":"2023-06-07T16:38:20.309804Z","iopub.status.idle":"2023-06-07T16:38:35.643643Z","shell.execute_reply.started":"2023-06-07T16:38:20.309742Z","shell.execute_reply":"2023-06-07T16:38:35.642154Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Step 2: Explore the training data\n\nWe'll start by loading a couple of training examples and using the IPython.display.Audio module to play them!","metadata":{}},{"cell_type":"code","source":"# Load a sample audio files from two different species\naudio_abe, sr_abe = librosa.load(\"/kaggle/input/birdclef-2023/train_audio/abethr1/XC128013.ogg\")\naudio_abh, sr_abh = librosa.load(\"/kaggle/input/birdclef-2023/train_audio/abhori1/XC127317.ogg\")","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:38:39.047249Z","iopub.execute_input":"2023-06-07T16:38:39.048263Z","iopub.status.idle":"2023-06-07T16:38:39.251884Z","shell.execute_reply.started":"2023-06-07T16:38:39.048210Z","shell.execute_reply":"2023-06-07T16:38:39.250379Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Play the audio\nAudio(data=audio_abe, rate=sr_abe)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Play the audio\nAudio(data=audio_abh, rate=sr_abh)","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Step 3: Match the model's output with the bird species in the competition\n\nThe competition includes 264 classes of birds, 261 of which exist in this model. We'll set up a way to map the model's output logits to our competition.","metadata":{}},{"cell_type":"code","source":"model = hub.load('https://kaggle.com/models/google/bird-vocalization-classifier/frameworks/tensorFlow2/variations/bird-vocalization-classifier/versions/1')\nlabels_path = hub.resolve('https://kaggle.com/models/google/bird-vocalization-classifier/frameworks/tensorFlow2/variations/bird-vocalization-classifier/versions/1') + \"/assets/label.csv\"","metadata":{"execution":{"iopub.status.busy":"2023-06-08T09:04:42.737480Z","iopub.execute_input":"2023-06-08T09:04:42.738532Z","iopub.status.idle":"2023-06-08T09:04:53.891679Z","shell.execute_reply.started":"2023-06-08T09:04:42.738482Z","shell.execute_reply":"2023-06-08T09:04:53.889676Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# Find the name of the class with the top score when mean-aggregated across frames.\ndef class_names_from_csv(class_map_csv_text):\n    \"\"\"Returns list of class names corresponding to score vector.\"\"\"\n    with open(labels_path) as csv_file:\n        csv_reader = csv.reader(csv_file, delimiter=',')\n        class_names = [mid for mid, desc in csv_reader]\n        return class_names[1:]\n\n## note that the bird classifier classifies a much larger set of birds than the\n## competition, so we need to load the model's set of class names or else our \n## indices will be off.\nclasses = class_names_from_csv(labels_path)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:39:18.199679Z","iopub.execute_input":"2023-06-07T16:39:18.200196Z","iopub.status.idle":"2023-06-07T16:39:18.225432Z","shell.execute_reply.started":"2023-06-07T16:39:18.200153Z","shell.execute_reply":"2023-06-07T16:39:18.223825Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"train_metadata = pd.read_csv(\"/kaggle/input/birdclef-2023/train_metadata.csv\")\ntrain_metadata.head()\ncompetition_classes = sorted(train_metadata.primary_label.unique())\n\nforced_defaults = 0\ncompetition_class_map = []\nfor c in competition_classes:\n    try:\n        i = classes.index(c)\n        competition_class_map.append(i)\n    except:\n        competition_class_map.append(0)\n        forced_defaults += 1\n        \n## this is the count of classes not supported by our pretrained model\n## you could choose to simply not predict these, set a default as above,\n## or create your own model using the pretrained model as a base.\nforced_defaults","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:39:25.603572Z","iopub.execute_input":"2023-06-07T16:39:25.604647Z","iopub.status.idle":"2023-06-07T16:39:25.825033Z","shell.execute_reply.started":"2023-06-07T16:39:25.604594Z","shell.execute_reply":"2023-06-07T16:39:25.823637Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"unsupported_classes = [competition_classes[i] for i in range(len(competition_classes)) if competition_class_map[i] == 0]\nprint(\"Class names not supported by the pretrained model:\")\nprint(unsupported_classes)\n","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:39:29.031737Z","iopub.execute_input":"2023-06-07T16:39:29.032223Z","iopub.status.idle":"2023-06-07T16:39:29.041387Z","shell.execute_reply.started":"2023-06-07T16:39:29.032165Z","shell.execute_reply":"2023-06-07T16:39:29.039689Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Step 4: Preprocess the data\n\nThe following functions are one way to load the audio provided and break it up into the five-second samples with a sample rate of 32,000 required by the competition.","metadata":{}},{"cell_type":"code","source":"def frame_audio(\n      audio_array: np.ndarray,\n      window_size_s: float = 5.0,\n      hop_size_s: float = 5.0,\n      sample_rate = 32000,\n      ) -> np.ndarray:\n    \n    \"\"\"Helper function for framing audio for inference.\"\"\"\n    \"\"\" using tf.signal \"\"\"\n    if window_size_s is None or window_size_s < 0:\n        return audio_array[np.newaxis, :]\n    frame_length = int(window_size_s * sample_rate)\n    hop_length = int(hop_size_s * sample_rate)\n    framed_audio = tf.signal.frame(audio_array, frame_length, hop_length, pad_end=True)\n    return framed_audio\n\ndef ensure_sample_rate(waveform, original_sample_rate,\n                       desired_sample_rate=32000):\n    \"\"\"Resample waveform if required.\"\"\"\n    if original_sample_rate != desired_sample_rate:\n        waveform = tfio.audio.resample(waveform, original_sample_rate, desired_sample_rate)\n    return desired_sample_rate, waveform","metadata":{"execution":{"iopub.status.busy":"2023-06-07T16:39:34.155844Z","iopub.execute_input":"2023-06-07T16:39:34.156356Z","iopub.status.idle":"2023-06-07T16:39:34.169530Z","shell.execute_reply.started":"2023-06-07T16:39:34.156304Z","shell.execute_reply":"2023-06-07T16:39:34.167642Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Below we load one training sample - use the Audio function to listen to the samples inside the notebook!","metadata":{}},{"cell_type":"code","source":"audio, sample_rate = librosa.load(\"/kaggle/input/dublinogg/Dublin_dataset_ogg/D3ECRI_0354.ogg\")\nsample_rate, wav_data = ensure_sample_rate(audio, sample_rate)\nAudio(wav_data, rate=sample_rate)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:22:45.774890Z","iopub.execute_input":"2023-06-07T17:22:45.775381Z","iopub.status.idle":"2023-06-07T17:22:46.076381Z","shell.execute_reply.started":"2023-06-07T17:22:45.775330Z","shell.execute_reply":"2023-06-07T17:22:46.075018Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Step 5: Make predictions\n\nEach test sample is cut into 5-second chunks. We use the pretrained model to return probabilities for all 10k birds included in the model, then pull out the classes used in this competition to create a final submission row. Note that we are NOT doing anything special to handle the 3 missing classes; those will need fine-tuning / transfer learning, which will be handled in a separate notebook.","metadata":{}},{"cell_type":"code","source":"fixed_tm = frame_audio(wav_data)\nlogits, embeddings = model.infer_tf(fixed_tm[:1])\nprobabilities = tf.nn.softmax(logits)\nargmax = np.argmax(probabilities)\nprint(f\"The audio is from the class {classes[argmax]} (element:{argmax} in the label.csv file), with probability of {probabilities[0][argmax]}\")","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:24:46.145506Z","iopub.execute_input":"2023-06-07T17:24:46.146864Z","iopub.status.idle":"2023-06-07T17:24:48.932198Z","shell.execute_reply.started":"2023-06-07T17:24:46.146796Z","shell.execute_reply":"2023-06-07T17:24:48.930742Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import os\nimport librosa\nimport numpy as np\nimport tensorflow as tf\nfrom IPython.display import Audio\n\n# Specify the directory path\ndirectory = \"/kaggle/input/dublinogg/Dublin_dataset_ogg\"\n\n# List all files in the directory\nfile_list = os.listdir(directory)\n\n# Iterate over each file in the directory\nfor file_name in file_list:\n    # Construct the full file path\n    file_path = os.path.join(directory, file_name)\n    \n    # Load the audio file\n    audio, sample_rate = librosa.load(file_path)\n    sample_rate, wav_data = ensure_sample_rate(audio, sample_rate)\n    \n    # Display the audio\n    Audio(wav_data, rate=sample_rate)\n    \n    # Perform further processing on the audio\n    fixed_tm = frame_audio(wav_data)\n    logits, embeddings = model.infer_tf(fixed_tm[:1])\n    probabilities = tf.nn.softmax(logits)\n    argmax = np.argmax(probabilities)\n    \n    # Print the results\n    print(f\"File: {file_name}\")\n    print(f\"The audio is from the class {classes[argmax]} (element: {argmax} in the label.csv file), with a probability of {probabilities[0][argmax]}\\n\")\n","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:26:48.563185Z","iopub.execute_input":"2023-06-07T17:26:48.565287Z","iopub.status.idle":"2023-06-07T17:27:44.502104Z","shell.execute_reply.started":"2023-06-07T17:26:48.565200Z","shell.execute_reply":"2023-06-07T17:27:44.500778Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import os\nimport librosa\nimport numpy as np\nimport tensorflow as tf\nimport json\nfrom IPython.display import Audio\n\n# Specify the directory path\ndirectory = \"/kaggle/input/dublinwav/Dublin_dataset_wav\"\n\n# List all files in the directory\nfile_list = os.listdir(directory)\n\n# Create a list to store all the predictions\npredictions = []\n\n# Iterate over each file in the directory\nfor file_name in file_list:\n    # Construct the full file path\n    file_path = os.path.join(directory, file_name)\n    \n    # Load the audio file\n    audio, sample_rate = librosa.load(file_path)\n    sample_rate, wav_data = ensure_sample_rate(audio, sample_rate)\n    \n    # Display the audio\n    Audio(wav_data, rate=sample_rate)\n    \n    # Perform further processing on the audio\n    fixed_tm = frame_audio(wav_data)\n    logits, embeddings = model.infer_tf(fixed_tm[:1])\n    probabilities = tf.nn.softmax(logits)\n    argmax = np.argmax(probabilities)\n    \n    # Store the prediction in a dictionary\n    prediction = {\n        \"file_name\": file_name,\n        \"class_name\": classes[argmax],\n        \"element\": int(argmax),\n        \"probability\": float(probabilities[0][argmax])\n    }\n    \n    # Append the prediction to the list\n    predictions.append(prediction)\n\n# Save the predictions in a text file\nwith open(\"predictions2.txt\", \"w\") as text_file:\n    for prediction in predictions:\n        text_file.write(f\"File: {prediction['file_name']}\\n\")\n        text_file.write(f\"The audio is from the class {prediction['class_name']} (element: {prediction['element']} in the label.csv file), with a probability of {prediction['probability']}\\n\\n\")\n\n# Save the predictions in a JSON file\nwith open(\"predictions2.json\", \"w\") as json_file:\n    json.dump(predictions, json_file, indent=4)\n","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:37:19.150056Z","iopub.execute_input":"2023-06-07T17:37:19.150548Z","iopub.status.idle":"2023-06-07T17:38:19.925569Z","shell.execute_reply.started":"2023-06-07T17:37:19.150507Z","shell.execute_reply":"2023-06-07T17:38:19.924099Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import os\nimport librosa\nimport numpy as np\nimport tensorflow as tf\nimport json\nfrom IPython.display import Audio\n\n# Specify the directory path\ndirectory = \"/kaggle/input/dublinwav/Dublin_dataset_wav\"\n\n# List all files in the directory with the \".wav\" extension\nfile_list = [file_name for file_name in os.listdir(directory) if file_name.endswith(\".wav\")]\n\n# Create a list to store all the predictions\npredictions = []\n\n# Iterate over each file in the directory\nfor file_name in file_list:\n    # Construct the full file path\n    file_path = os.path.join(directory, file_name)\n    \n    # Load the audio file\n    wav_data, sample_rate = librosa.load(file_path, sr=None, mono=True)\n    \n    # Display the audio\n    Audio(wav_data, rate=sample_rate)\n    \n    # Perform further processing on the audio\n    fixed_tm = frame_audio(wav_data)\n    logits, embeddings = model.infer_tf(fixed_tm[:1])\n    probabilities = tf.nn.softmax(logits)\n    argmax = np.argmax(probabilities)\n    \n    # Store the prediction in a dictionary\n    prediction = {\n        \"file_name\": file_name,\n        \"class_name\": classes[argmax],\n        \"element\": int(argmax),\n        \"probability\": float(probabilities[0][argmax])\n    }\n    \n    # Append the prediction to the list\n    predictions.append(prediction)\n\n# Save the predictions in a text file\nwith open(\"predictions.txt\", \"w\") as text_file:\n    for prediction in predictions:\n        text_file.write(f\"File: {prediction['file_name']}\\n\")\n        text_file.write(f\"The audio is from the class {prediction['class_name']} (element: {prediction['element']} in the label.csv file), with a probability of {prediction['probability']}\\n\\n\")\n\n# Save the predictions in a JSON file\nwith open(\"predictions.json\", \"w\") as json_file:\n    json.dump(predictions, json_file, indent=4)\n","metadata":{"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def predict_for_sample(filename, sample_submission, frame_limit_secs=None):\n    file_id = filename.split(\".ogg\")[0].split(\"/\")[-1]\n    \n    audio, sample_rate = librosa.load(filename)\n    sample_rate, wav_data = ensure_sample_rate(audio, sample_rate)\n    \n    fixed_tm = frame_audio(wav_data)\n    \n    frame = 5\n    all_logits, all_embeddings = model.infer_tf(fixed_tm[:1])\n    for window in fixed_tm[1:]:\n        if frame_limit_secs and frame > frame_limit_secs:\n            continue\n        \n        logits, embeddings = model.infer_tf(window[np.newaxis, :])\n        all_logits = np.concatenate([all_logits, logits], axis=0)\n        frame += 5\n    \n    frame = 5\n    all_probabilities = []\n    for frame_logits in all_logits:\n        probabilities = tf.nn.softmax(frame_logits).numpy()\n        \n        ## set the appropriate row in the sample submission\n        sample_submission.loc[sample_submission.row_id == file_id + \"_\" + str(frame), competition_classes] = probabilities[competition_class_map]\n        frame += 5","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:23:36.166829Z","iopub.execute_input":"2023-06-07T17:23:36.167359Z","iopub.status.idle":"2023-06-07T17:23:36.183146Z","shell.execute_reply.started":"2023-06-07T17:23:36.167312Z","shell.execute_reply":"2023-06-07T17:23:36.181537Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## Step 6: Generate a submission\n\nNow we process all of the test samples as discussed above, creating output rows, and saving them in the provided `sample_submission.csv`. Finally, we save these rows to our final output file: `submission.csv`. This is the file that gets submitted and scored when you submit the notebook.","metadata":{}},{"cell_type":"code","source":"test_samples = list(glob.glob(\"/kaggle/input/dublinogg/*.ogg\"))\ntest_samples","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:24:10.959363Z","iopub.execute_input":"2023-06-07T17:24:10.959816Z","iopub.status.idle":"2023-06-07T17:24:10.972491Z","shell.execute_reply.started":"2023-06-07T17:24:10.959777Z","shell.execute_reply":"2023-06-07T17:24:10.970988Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample_sub = pd.read_csv(\"/kaggle/input/birdclef-2023/sample_submission.csv\")\nsample_sub[competition_classes] = sample_sub[competition_classes].astype(np.float32)\nsample_sub.head()","metadata":{"execution":{"iopub.status.busy":"2023-06-07T17:24:27.304427Z","iopub.execute_input":"2023-06-07T17:24:27.304866Z","iopub.status.idle":"2023-06-07T17:24:27.462058Z","shell.execute_reply.started":"2023-06-07T17:24:27.304829Z","shell.execute_reply":"2023-06-07T17:24:27.460662Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"frame_limit_secs = 15 if sample_sub.shape[0] == 3 else None\nfor sample_filename in test_samples:\n    predict_for_sample(sample_filename, sample_sub, frame_limit_secs=15)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T10:31:43.877108Z","iopub.execute_input":"2023-06-07T10:31:43.877473Z","iopub.status.idle":"2023-06-07T10:31:48.136029Z","shell.execute_reply.started":"2023-06-07T10:31:43.877438Z","shell.execute_reply":"2023-06-07T10:31:48.135117Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample_sub","metadata":{"execution":{"iopub.status.busy":"2023-06-07T10:31:54.588090Z","iopub.execute_input":"2023-06-07T10:31:54.588458Z","iopub.status.idle":"2023-06-07T10:31:54.619635Z","shell.execute_reply.started":"2023-06-07T10:31:54.588429Z","shell.execute_reply":"2023-06-07T10:31:54.618250Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"sample_sub.to_csv(\"submission.csv\", index=False)","metadata":{"execution":{"iopub.status.busy":"2023-06-07T10:32:03.381443Z","iopub.execute_input":"2023-06-07T10:32:03.381779Z","iopub.status.idle":"2023-06-07T10:32:03.401038Z","shell.execute_reply.started":"2023-06-07T10:32:03.381752Z","shell.execute_reply":"2023-06-07T10:32:03.399077Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}