[ADD] Folder containing cloud_analysis scripts

cloud_analysis/Dockerfile

@@ -0,0 +1,33 @@
+FROM python:3.8-bullseye
+
+ARG PACKAGES="ffmpeg build-essential unzip gcsfuse"
+
+ARG DEBIAN_FRONTEND=noninteractive
+RUN \
+    echo "deb https://packages.cloud.google.com/apt gcsfuse-bullseye main" | tee /etc/apt/sources.list.d/gcsfuse.list && \
+    curl https://packages.cloud.google.com/apt/doc/apt-key.gpg | apt-key add - && \
+    apt-get update && \
+    apt-get install -qq $PACKAGES && \
+    rm -rf /var/lib/apt/lists/*
+
+RUN pip3 install poetry 
+
+WORKDIR /app
+
+# Clone the repository and download assets
+RUN mkdir audioclip
+RUN cd audioclip 
+RUN wget https://zenodo.org/record/7969521/files/assets.zip?download=1 
+RUN unzip ./assets.zip?download=1 
+RUN cd ../
+
+# Package install
+COPY . ./
+RUN poetry config virtualenvs.create false
+RUN poetry install --no-root
+
+
+
+ENV PYTHONPATH "${PYTHONPATH}:/app:/app/audioclip:/app:/app/src"
+ENTRYPOINT ["python", "/app/cloud_analysis/main.py"]
+

cloud_analysis/instructions.md

@@ -0,0 +1,64 @@
+# Documentation on running a Docker container on google cloud
+
+## General stuff with GCloud UI
+
+- If lost, there is a button `console` on the upper right corner that brings us back to the projects
+
+
+## Login stuff
+
+1- Generate the Google Cloud credentials:
+
+```
+gcloud auth application-default login
+```
+
+The credentials are stored in `~/.config/gcloud/`
+
+2- Login to your Google Cloud account:
+
+```
+gcloud auth login
+```
+
+3- Log in to the project
+
+```
+gcloud config set project snokuter-akustikk
+```
+
+## Docker stuff
+
+1- Enable services
+
+Enable the Cloud Functions, Cloud Run, and Cloud Build APIs:
+
+```
+gcloud services enable cloudfunctions.googleapis.com
+gcloud services enable run.googleapis.com
+gcloud services enable cloudbuild.googleapis.com
+```
+
+2- Set up credentials on Google Cloud:
+
+The credentials are stored in `~/.docker/config.json`
+
+```
+gcloud auth configure-docker
+```
+
+3- Change the name of the image to push it on GCloud
+
+The Docker image will be saved in Google's container registry, which is directly supported by Cloud Run
+
+```
+docker tag ghcr.io/ninanor/snowmobile_analyzer:main gcr.io/snoskuter-akustikk/snowmobile_analyzer:main
+docker push gcr.io/snoskuter-akustikk/snowmobile_analyzer:main
+
+```
+
+4- Deploy the Docker image to Google Cloud run
+
+```
+gcloud run deploy model --image gcr.io/snoskuter-akustikk/snowmobile_analyzer:main
+```

cloud_analysis/main.py

@@ -0,0 +1,119 @@
+
+#!/usr/env/bin python3
+
+from flask import Flask, request, jsonify
+app = Flask(__name__)
+
+import argparse
+import numpy as np
+
+import torch
+from torch.utils.data import DataLoader
+
+from predict import predict
+from predict import initModel
+from utils.utils import AudioList
+
+
+def analyseAudioFile(
+        audio_file_path, batch_size=1, num_workers=4, min_hr = 0.1, min_conf = 0.99
+):
+    # Initiate model
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    model_path = "./assets/snowmobile_model.pth"
+    model = initModel(model_path=model_path, device=device)
+
+    # Run the predictions
+    list_preds = AudioList().get_processed_list(audio_file_path)
+    predLoader = DataLoader(list_preds, batch_size=batch_size, num_workers=num_workers, pin_memory=False)
+    prob_audioclip_array, hr_array = predict(predLoader, model, device)
+
+    # Set index of the array to 0 (beginning)
+    idx_begin = 0
+
+    # List of results
+    results = []
+
+    for item_audioclip, item_hr in zip(prob_audioclip_array, hr_array):
+
+        # Get the properties of the detection (start, end, label, confidence and harmonic ratio)
+        idx_end = idx_begin + 3
+        conf = np.array(item_audioclip)
+        label = np.argmax(conf, axis=0)
+        confidence = conf.max()
+        hr = np.array(item_hr)
+
+        # If the label is not "soundscape" then write the row:
+        if label != 0 and hr > min_hr and confidence > min_conf:
+            item_properties = [idx_begin, idx_end, confidence, hr]
+            results.append(item_properties)
+
+        # Update the start time of the detection
+        idx_begin = idx_end
+
+    return results
+
+def on_process_audio(
+        audio_id: str, audio_rec: dict, audio_file_path: str
+):
+
+    print(f"PROCESSING audioId={audio_id}")
+    location = audio_rec["location"]
+
+    # A call out to your code here. Optionally we can pass on the recorder coordinates 
+    results = analyseAudioFile(audio_file_path)
+    # The object results is a list containing detections in the form:
+    # [start, end, confidence, harmonic ratio]
+
+    # After processing we record each detection in the database. 
+    # Each detection should have a start and end time which is used to create an audio clip later. 
+    detections = []
+    for r in results: 
+        start, end, confidence, harmonic_ratio = r
+        # create the detections
+        detections.append({
+            u"start": start,
+            u"end": end,
+            u"tags": ["snowmobile"],
+            #u"analysisId": analysis_id,
+            # Add any other information you want to record here
+            u"confidence": confidence,
+            u"harmonic_ratio": harmonic_ratio
+        })
+
+    print(detections)
+    print(f"{audio_id} completed with {len(detections)} detections")
+
+
+@app.route('/process-audio', methods=['POST'])
+def process_audio_endpoint():
+    audio_file_path = request.json['audio_file_path']
+    audio_id = request.json['audio_id']
+    audio_rec = request.json['audio_rec']
+
+    on_process_audio(audio_id, audio_rec, audio_file_path)
+
+    return jsonify({"message": "Audio processing completed!"})
+
+
+if __name__ == "__main__":
+
+
+    parser = argparse.ArgumentParser()
+
+    parser.add_argument(
+        "--input",
+        help="Path to the file to analyze",
+        required=True,
+        type=str,
+    )
+
+    cli_args = parser.parse_args()
+
+    on_process_audio("my-audio-id", {
+        "location": {
+            "latitude": 0,
+            "longitude": 0
+        }
+    }, cli_args.input)
+