add conv-autoencoder architecture

dcase2020_task2/data_sets/mcm_dataset.py

@@ -55,7 +55,8 @@ class MCMDataSet(BaseDataSet):
             power=1.0,
             fmin=0,
             normalize_raw=False,
-            normalize=None
+            normalize=None,
+            hop_all=False
     ):
         self.data_root = data_root
         self.context = context
@@ -65,6 +66,7 @@ class MCMDataSet(BaseDataSet):
         self.power = power
         self.fmin = fmin
         self.normalize = normalize
+        self.hop_all = hop_all
 
         assert type(machine_type) == int and type(machine_id) == int
 
@@ -76,7 +78,8 @@ class MCMDataSet(BaseDataSet):
             'hop_size': self.hop_size,
             'power': power,
             'normalize': normalize_raw,
-            'fmin': fmin
+            'fmin': fmin,
+            'hop_all': hop_all
         }
 
         training_set = MachineDataSet(machine_type, machine_id, mode='training', **kwargs)
@@ -127,7 +130,8 @@ class MachineDataSet(torch.utils.data.Dataset):
             hop_size=512,
             power=2.0,
             normalize=True,
-            fmin=0
+            fmin=0,
+            hop_all=False
     ):
 
         assert mode in ['training', 'validation', 'testing']
@@ -143,6 +147,7 @@ class MachineDataSet(torch.utils.data.Dataset):
         self.machine_type = INVERSE_CLASS_MAP[machine_type]
         self.machine_id = machine_id
         self.fmin = fmin
+        self.hop_all = hop_all
 
         if mode == 'training':
             files = glob.glob(
@@ -165,7 +170,7 @@ class MachineDataSet(torch.utils.data.Dataset):
         files = sorted(files)
         self.files = files
         self.file_length = self.__load_preprocess_file__(files[0]).shape[-1]
-        self.num_samples_per_file = self.file_length - self.context + 1
+        self.num_samples_per_file = (self.file_length // self.context) if hop_all else (self.file_length - self.context + 1)
         self.meta_data = self.__load_meta_data__(files)
         self.data = self.__load_data__(files)
 
@@ -175,7 +180,7 @@ class MachineDataSet(torch.utils.data.Dataset):
         # get audio file index
         item = item // self.num_samples_per_file
         # load audio file and extract audio junk
-        offset = item * self.file_length + offset
+        offset = item * self.file_length + ((offset * self.context) if self.hop_all else offset)
         observation = self.data[:, offset:offset + self.context]
         # create data object
         meta_data = self.meta_data[item].copy()

dcase2020_task2/experiments/baseline_experiment.py

@@ -156,7 +156,8 @@ def configuration():
             'hop_size': hop_size,
             'normalize_raw': normalize_raw,
             'power': power,
-            'fmin': fmin
+            'fmin': fmin,
+            'hop_all': True
         }
     }