diff --git a/.gitignore b/.gitignore
index 909b4a7..7906666 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,3 +1,9 @@
+# IDES
+*.wpr
+*.wpu
+*.udb
+*.ann
+
 # Byte-compiled / optimized / DLL files
 __pycache__/
 *.py[cod]
diff --git a/parakeet/data/datasets.py b/parakeet/data/datasets.py
new file mode 100644
index 0000000..023577d
--- /dev/null
+++ b/parakeet/data/datasets.py
@@ -0,0 +1,22 @@
+from paddle.io import Dataset
+
+from os import listdir
+from os.path import splitext, join
+import librosa
+
+class AudioFolderDataset(Dataset):
+    def __init__(self, path, sample_rate, extension="wav"):
+        self.root = path
+        self.sample_rate = sample_rate
+        self.extension = extension
+        self.file_names = [join(self.root, x) for x in listdir(self.root) \
+            if splitext(x)[-1] == self.extension]
+        self.length = len(self.file_names)
+
+    def __len__(self):
+        return self.length
+
+    def __getitem__(self, i):
+        file_name = self.file_names[i]
+        y, sr = librosa.load(file_name, sr=self.sample_rate) # pylint: disable=unused-variable
+        return y
\ No newline at end of file