Fix clip missalignment in validation dataset

src/batdetect2/train/callbacks.py

@@ -126,7 +126,7 @@ class ValidationMetrics(Callback):
         dataset = self.get_dataset(trainer)
 
         clip_annotations = [
-            dataset.clip_annotations[int(example_idx)]
+            dataset.get_clip_annotation(int(example_idx))
             for example_idx in batch.idx
         ]
 

src/batdetect2/train/clips.py

@@ -158,8 +158,8 @@ class PaddedClipConfig(BaseConfig):
 
 @registry.register(PaddedClipConfig)
 class PaddedClip:
-    def __init__(self, duration: float = DEFAULT_TRAIN_CLIP_DURATION):
+    def __init__(self, chunk_size: float = DEFAULT_TRAIN_CLIP_DURATION):
-        self.duration = duration
+        self.chunk_size = chunk_size
 
     def __call__(
         self,
@@ -168,7 +168,9 @@ class PaddedClip:
         clip = clip_annotation.clip
         duration = clip.duration
 
-        target_duration = self.duration * np.ceil(duration / self.duration)
+        target_duration = float(
+            self.chunk_size * np.ceil(duration / self.chunk_size)
+        )
         clip = clip.model_copy(
             update=dict(
                 end_time=clip.start_time + target_duration,
@@ -178,7 +180,7 @@ class PaddedClip:
 
     @classmethod
     def from_config(cls, config: PaddedClipConfig):
-        return cls(duration=config.chunk_size)
+        return cls(chunk_size=config.chunk_size)
 
 
 ClipConfig = Annotated[

src/batdetect2/train/dataset.py

@@ -1,4 +1,4 @@
-from typing import Optional, Sequence
+from typing import Optional, Sequence, Tuple
 
 import torch
 from soundevent import data
@@ -101,20 +101,10 @@ class ValidationDataset(Dataset):
         return len(self.clip_annotations)
 
     def __getitem__(self, idx) -> TrainExample:
-        clip_annotation = self.clip_annotations[idx]
+        wav, clip_annotation = self.load_audio(idx)
-
-        if self.clipper is not None:
-            clip_annotation = self.clipper(clip_annotation)
-
         clip = clip_annotation.clip
-        wav = self.audio_loader.load_clip(
-            clip_annotation.clip,
-            audio_dir=self.audio_dir,
-        )
 
-        wav_tensor = torch.tensor(wav).unsqueeze(0)
+        spectrogram = self.preprocessor(wav)
-
-        spectrogram = self.preprocessor(wav_tensor)
 
         heatmaps = self.labeller(clip_annotation, spectrogram)
 
@@ -127,3 +117,17 @@ class ValidationDataset(Dataset):
             start_time=torch.tensor(clip.start_time),
             end_time=torch.tensor(clip.end_time),
         )
+
+    def get_clip_annotation(self, idx: int) -> data.ClipAnnotation:
+        clip_annotation = self.clip_annotations[idx]
+
+        if self.clipper is not None:
+            clip_annotation = self.clipper(clip_annotation)
+
+        return clip_annotation
+
+    def load_audio(self, idx: int) -> Tuple[torch.Tensor, data.ClipAnnotation]:
+        clip_annotation = self.get_clip_annotation(idx)
+        clip = clip_annotation.clip
+        wav = self.audio_loader.load_clip(clip, audio_dir=self.audio_dir)
+        return torch.tensor(wav).unsqueeze(0), clip_annotation