add places365 dataset

tensorpack/dataflow/dataset/ilsvrc.py

@@ -192,8 +192,7 @@ def __init__(self, dir, name, meta_dir=None,
             dir (str): A directory containing a subdir named ``name``,
                 containing the images in a structure described below.
             name (str): One of 'train' or 'val' or 'test'.
-            shuffle (bool): shuffle the dataset.
-                Defaults to True if name=='train'.
+            shuffle (bool): shuffle the dataset. Defaults to True if name=='train'.
             dir_structure (str): One of 'original' or 'train'.
                 The directory structure for the 'val' directory.
                 'original' means the original decompressed directory, which only has list of image files (as below).
@@ -354,6 +353,7 @@ def __iter__(self):
 except ImportError:
     from ...utils.develop import create_dummy_class
     ILSVRC12 = create_dummy_class('ILSVRC12', 'cv2')  # noqa
+    TinyImageNet = create_dummy_class('TinyImageNet', 'cv2')  # noqa
 
 if __name__ == '__main__':
     meta = ILSVRCMeta()

tensorpack/dataflow/dataset/places.py

@@ -0,0 +1,81 @@
+#-*- coding: utf-8 -*-
+
+import os
+import numpy as np
+
+from ...utils import logger
+from ..base import RNGDataFlow
+
+
+class Places365Standard(RNGDataFlow):
+    """
+    The Places365-Standard Dataset, in low resolution format only.
+    Produces BGR images of shape (256, 256, 3) in range [0, 255].
+    """
+    def __init__(self, dir, name, shuffle=None):
+        """
+        Args:
+            dir: path to the Places365-Standard dataset in its "easy directory
+                structure".  See http://places2.csail.mit.edu/download.html
+            name: one of "train" or "val"
+            shuffle (bool): shuffle the dataset. Defaults to True if name=='train'.
+        """
+        assert name in ['train', 'val'], name
+        dir = os.path.expanduser(dir)
+        assert os.path.isdir(dir), dir
+        self.name = name
+        if shuffle is None:
+            shuffle = name == 'train'
+        self.shuffle = shuffle
+
+        label_file = os.path.join(dir, name + ".txt")
+        all_files = []
+        labels = set()
+        with open(label_file) as f:
+            for line in f:
+                filepath = os.path.join(dir, line.strip())
+                line = line.strip().split("/")
+                label = line[1]
+                all_files.append((filepath, label))
+                labels.add(label)
+        self._labels = sorted(list(labels))
+        # class ids are sorted alphabetically:
+        # https://github.com/CSAILVision/places365/blob/master/categories_places365.txt
+        labelmap = {label: id for id, label in enumerate(self._labels)}
+        self._files = [(path, labelmap[x]) for path, x in all_files]
+        logger.info("Found {} images in {}.".format(len(self._files), label_file))
+
+    def get_label_names(self):
+        """
+        Returns:
+            [str]: name of each class.
+        """
+        return self._labels
+
+    def __len__(self):
+        return len(self._files)
+
+    def __iter__(self):
+        idxs = np.arange(len(self._files))
+        if self.shuffle:
+            self.rng.shuffle(idxs)
+        for k in idxs:
+            fname, label = self._files[k]
+            im = cv2.imread(fname, cv2.IMREAD_COLOR)
+            assert im is not None, fname
+            yield [im, label]
+
+
+try:
+    import cv2
+except ImportError:
+    from ...utils.develop import create_dummy_class
+    Places365Standard = create_dummy_class('Places365Standard', 'cv2')  # noqa
+
+if __name__ == '__main__':
+    from tensorpack.dataflow import PrintData
+    ds = Places365Standard("~/data/places365_standard/", 'train')
+    ds = PrintData(ds, num=100)
+    ds.reset_state()
+    for k in ds:
+        pass