Dev/spectrogram updates

.gitattributes

@@ -1 +1,2 @@
 assets/example.wav filter=lfs diff=lfs merge=lfs -text
+assets/model.mobilenet.onnx filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -2,3 +2,10 @@
 /**/*.shp
 /**/*.shx
 /**/*.csv
+models/datasets/
+models/spectrograms/
+models/ignore/
+models/*.jpg
+models/*.pkl
+temp*.jpg
+temp*.png

README.md

@@ -15,8 +15,9 @@ This is the simplest configuration for developers to start with.
 
 ### Run Vue Frontend
 
-1. Run `npm install`
-2. Run `npm run dev`
+1. Run `cd client/`
+2. Run `npm install`
+3. Run `npm run dev`
 
 ### Run Application
 

bats_ai/core/admin/__init__.py

@@ -1,4 +1,5 @@
 from .annotations import AnnotationsAdmin
+from .compressed_spectrogram import CompressedSpectrogramAdmin
 from .grts_cells import GRTSCellsAdmin
 from .image import ImageAdmin
 from .recording import RecordingAdmin
@@ -14,4 +15,5 @@
     'TemporalAnnotationsAdmin',
     'SpeciesAdmin',
     'GRTSCellsAdmin',
+    'CompressedSpectrogramAdmin',
 ]

bats_ai/core/admin/compressed_spectrogram.py

@@ -0,0 +1,30 @@
+from django.contrib import admin
+
+from bats_ai.core.models import CompressedSpectrogram
+
+
+@admin.register(CompressedSpectrogram)
+class CompressedSpectrogramAdmin(admin.ModelAdmin):
+    list_display = [
+        'pk',
+        'recording',
+        'spectrogram',
+        'image_file',
+        'length',
+        'widths',
+        'starts',
+        'stops',
+    ]
+    list_display_links = ['pk', 'recording', 'spectrogram']
+    list_select_related = True
+    autocomplete_fields = ['recording']
+    readonly_fields = [
+        'recording',
+        'spectrogram',
+        'image_file',
+        'created',
+        'modified',
+        'widths',
+        'starts',
+        'stops',
+    ]

bats_ai/core/admin/recording.py

@@ -64,7 +64,6 @@ def spectrogram_status(self, recording: Recording):
     def compute_spectrograms(self, request: HttpRequest, queryset: QuerySet):
         counter = 0
         for recording in queryset:
-            if not recording.has_spectrogram:
-                recording_compute_spectrogram.delay(recording.pk)
-                counter += 1
+            recording_compute_spectrogram.delay(recording.pk)
+            counter += 1
         self.message_user(request, f'{counter} recordings queued', messages.SUCCESS)

bats_ai/core/admin/spectrogram.py

@@ -8,6 +8,7 @@ class SpectrogramAdmin(admin.ModelAdmin):
     list_display = [
         'pk',
         'recording',
+        'colormap',
         'image_file',
         'width',
         'height',
@@ -20,6 +21,7 @@ class SpectrogramAdmin(admin.ModelAdmin):
     autocomplete_fields = ['recording']
     readonly_fields = [
         'recording',
+        'colormap',
         'image_file',
         'created',
         'modified',

bats_ai/core/migrations/0010_recording_computed_species_recording_detector_and_more.py → bats_ai/core/migrations/0009_annotations_type_recording_computed_species_and_more.py

@@ -1,14 +1,19 @@
-# Generated by Django 4.1.13 on 2024-04-03 13:07
+# Generated by Django 4.1.13 on 2024-04-11 13:06
 
 from django.db import migrations, models
 
 
 class Migration(migrations.Migration):
     dependencies = [
-        ('core', '0009_annotations_type'),
+        ('core', '0008_grtscells_recording_recorded_time'),
     ]
 
     operations = [
+        migrations.AddField(
+            model_name='annotations',
+            name='type',
+            field=models.TextField(blank=True, null=True),
+        ),
         migrations.AddField(
             model_name='recording',
             name='computed_species',
@@ -48,4 +53,9 @@ class Migration(migrations.Migration):
             name='unusual_occurrences',
             field=models.TextField(blank=True, null=True),
         ),
+        migrations.AddField(
+            model_name='spectrogram',
+            name='colormap',
+            field=models.CharField(max_length=20, null=True),
+        ),
     ]

bats_ai/core/migrations/0010_compressedspectrogram.py

@@ -0,0 +1,84 @@
+# Generated by Django 4.1.13 on 2024-04-19 13:55
+
+import django.contrib.postgres.fields
+from django.db import migrations, models
+import django.db.models.deletion
+import django_extensions.db.fields
+
+
+class Migration(migrations.Migration):
+    dependencies = [
+        ('core', '0009_annotations_type_recording_computed_species_and_more'),
+    ]
+
+    operations = [
+        migrations.CreateModel(
+            name='CompressedSpectrogram',
+            fields=[
+                (
+                    'id',
+                    models.BigAutoField(
+                        auto_created=True, primary_key=True, serialize=False, verbose_name='ID'
+                    ),
+                ),
+                (
+                    'created',
+                    django_extensions.db.fields.CreationDateTimeField(
+                        auto_now_add=True, verbose_name='created'
+                    ),
+                ),
+                (
+                    'modified',
+                    django_extensions.db.fields.ModificationDateTimeField(
+                        auto_now=True, verbose_name='modified'
+                    ),
+                ),
+                ('image_file', models.FileField(upload_to='')),
+                ('length', models.IntegerField()),
+                (
+                    'starts',
+                    django.contrib.postgres.fields.ArrayField(
+                        base_field=django.contrib.postgres.fields.ArrayField(
+                            base_field=models.IntegerField(), size=None
+                        ),
+                        size=None,
+                    ),
+                ),
+                (
+                    'stops',
+                    django.contrib.postgres.fields.ArrayField(
+                        base_field=django.contrib.postgres.fields.ArrayField(
+                            base_field=models.IntegerField(), size=None
+                        ),
+                        size=None,
+                    ),
+                ),
+                (
+                    'widths',
+                    django.contrib.postgres.fields.ArrayField(
+                        base_field=django.contrib.postgres.fields.ArrayField(
+                            base_field=models.IntegerField(), size=None
+                        ),
+                        size=None,
+                    ),
+                ),
+                ('cache_invalidated', models.BooleanField(default=True)),
+                (
+                    'recording',
+                    models.ForeignKey(
+                        on_delete=django.db.models.deletion.CASCADE, to='core.recording'
+                    ),
+                ),
+                (
+                    'spectrogram',
+                    models.ForeignKey(
+                        on_delete=django.db.models.deletion.CASCADE, to='core.spectrogram'
+                    ),
+                ),
+            ],
+            options={
+                'get_latest_by': 'modified',
+                'abstract': False,
+            },
+        ),
+    ]

bats_ai/core/models/__init__.py

@@ -1,7 +1,8 @@
 from .annotations import Annotations
+from .compressed_spectrogram import CompressedSpectrogram
 from .grts_cells import GRTSCells
 from .image import Image
-from .recording import Recording
+from .recording import Recording, colormap
 from .recording_annotation_status import RecordingAnnotationStatus
 from .species import Species
 from .spectrogram import Spectrogram
@@ -16,4 +17,6 @@
     'Spectrogram',
     'TemporalAnnotations',
     'GRTSCells',
+    'colormap',
+    'CompressedSpectrogram',
 ]

bats_ai/core/models/compressed_spectrogram.py

@@ -0,0 +1,112 @@
+from PIL import Image
+import cv2
+from django.contrib.postgres.fields import ArrayField
+from django.core.files.storage import default_storage
+from django.db import models
+from django.dispatch import receiver
+from django_extensions.db.models import TimeStampedModel
+import numpy as np
+
+from .recording import Recording
+from .spectrogram import Spectrogram
+
+FREQ_MIN = 5e3
+FREQ_MAX = 120e3
+FREQ_PAD = 2e3
+
+
+# TimeStampedModel also provides "created" and "modified" fields
+class CompressedSpectrogram(TimeStampedModel, models.Model):
+    recording = models.ForeignKey(Recording, on_delete=models.CASCADE)
+    spectrogram = models.ForeignKey(Spectrogram, on_delete=models.CASCADE)
+    image_file = models.FileField()
+    length = models.IntegerField()
+    starts = ArrayField(ArrayField(models.IntegerField()))
+    stops = ArrayField(ArrayField(models.IntegerField()))
+    widths = ArrayField(ArrayField(models.IntegerField()))
+    cache_invalidated = models.BooleanField(default=True)
+
+    @property
+    def image_url(self):
+        return default_storage.url(self.image_file.name)
+
+    def predict(self):
+        import json
+        import os
+
+        import onnx
+        import onnxruntime as ort
+        import tqdm
+
+        img = Image.open(self.image_file)
+
+        relative = ('..',) * 4
+        asset_path = os.path.abspath(os.path.join(__file__, *relative, 'assets'))
+
+        onnx_filename = os.path.join(asset_path, 'model.mobilenet.onnx')
+        assert os.path.exists(onnx_filename)
+
+        session = ort.InferenceSession(
+            onnx_filename,
+            providers=[
+                (
+                    'CUDAExecutionProvider',
+                    {
+                        'cudnn_conv_use_max_workspace': '1',
+                        'device_id': 0,
+                        'cudnn_conv_algo_search': 'HEURISTIC',
+                    },
+                ),
+                'CPUExecutionProvider',
+            ],
+        )
+
+        img = np.array(img)
+
+        h, w, c = img.shape
+        ratio_y = 224 / h
+        ratio_x = ratio_y * 0.5
+        raw = cv2.resize(img, None, fx=ratio_x, fy=ratio_y, interpolation=cv2.INTER_LANCZOS4)
+
+        h, w, c = raw.shape
+        if w <= h:
+            canvas = np.zeros((h, h + 1, 3), dtype=raw.dtype)
+            canvas[:, :w, :] = raw
+            raw = canvas
+            h, w, c = raw.shape
+
+        inputs_ = []
+        for index in range(0, w - h, 100):
+            inputs_.append(raw[:, index : index + h, :])
+        inputs_.append(raw[:, -h:, :])
+        inputs_ = np.array(inputs_)
+
+        chunksize = 1
+        chunks = np.array_split(inputs_, np.arange(chunksize, len(inputs_), chunksize))
+        outputs = []
+        for chunk in tqdm.tqdm(chunks, desc='Inference'):
+            outputs_ = session.run(
+                None,
+                {'input': chunk},
+            )
+            outputs.append(outputs_[0])
+        outputs = np.vstack(outputs)
+        outputs = outputs.mean(axis=0)
+
+        model = onnx.load(onnx_filename)
+        mapping = json.loads(model.metadata_props[0].value)
+        labels = [mapping['forward'][str(index)] for index in range(len(mapping['forward']))]
+
+        prediction = np.argmax(outputs)
+        label = labels[prediction]
+        score = outputs[prediction]
+
+        confs = dict(zip(labels, outputs))
+
+        return label, score, confs
+
+
+@receiver(models.signals.pre_delete, sender=Spectrogram)
+def delete_content(sender, instance, **kwargs):
+    if instance.image_file:
+        instance.image_file.delete(save=False)