Merge branch 'master' into train-ssd

chainercv/datasets/voc/voc_detection_dataset.py

@@ -29,7 +29,7 @@ class VOCDetectionDataset(chainer.dataset.DatasetMixin):
     :math:`(R, 4)`, where :math:`R` is the number of bounding boxes in
     the image. The second axis represents attributes of the bounding box.
     They are :obj:`(y_min, x_min, y_max, x_max)`, where the
-    four attributes are coordinates of the bottom left and the top right
+    four attributes are coordinates of the top left and the bottom right
     vertices.
 
     The labels are packed into a one dimensional tensor of shape :math:`(R,)`.

chainercv/links/model/segnet/segnet_basic.py

@@ -176,6 +176,6 @@ def predict(self, imgs):
                 dtype = score.dtype
                 score = resize(score, (H, W)).astype(dtype)
 
-            label = np.argmax(score, axis=0)
+            label = np.argmax(score, axis=0).astype(np.int32)
             labels.append(label)
         return labels

chainercv/links/model/ssd/multibox_coder.py

@@ -231,8 +231,8 @@ def decode(self, mb_loc, mb_conf, nms_thresh, score_thresh):
             label.append(xp.array((l,) * len(bbox_l)))
             score.append(score_l)
 
-        bbox = xp.vstack(bbox)
-        label = xp.hstack(label).astype(int)
-        score = xp.hstack(score)
+        bbox = xp.vstack(bbox).astype(np.float32)
+        label = xp.hstack(label).astype(np.int32)
+        score = xp.hstack(score).astype(np.float32)
 
         return bbox, label, score

chainercv/transforms/bbox/crop_bbox.py

@@ -4,29 +4,29 @@
 def crop_bbox(
         bbox, y_slice=None, x_slice=None,
         allow_outside_center=True, return_param=False):
-    """Crop bounding boxes.
+    """Translate bounding boxes to fit within the cropped area of an image.
 
     This method is mainly used together with image cropping.
     This method translates the coordinates of bounding boxes like
     :func:`~chainercv.transforms.translate_bbox`. In addition,
-    this function truncates the bounding boxes to fit within the cropping area.
-    If a bounding box does not overlap with the cropping area,
+    this function truncates the bounding boxes to fit within the cropped area.
+    If a bounding box does not overlap with the cropped area,
     this bounding box will be removed.
 
     The bounding boxes are expected to be packed into a two dimensional
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
     the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
-    where the four attributes are coordinates of the bottom left and the
-    top right vertices.
+    where the four attributes are coordinates of the top left and the
+    bottom right vertices.
 
     Args:
         bbox (~numpy.ndarray): Bounding boxes to be transformed. The shape is
             :math:`(R, 4)`. :math:`R` is the number of bounding boxes.
         y_slice (slice): The slice of y axis.
         x_slice (slice): The slice of x axis.
         allow_outside_center (bool): If this argument is :obj:`False`,
-            bounding boxes whose centers are outside of the cropping area
+            bounding boxes whose centers are outside of the cropped area
             are removed. The default value is :obj:`True`.
         return_param (bool): If :obj:`True`, this function returns
             indices of kept bounding boxes.

chainercv/transforms/bbox/flip_bbox.py

@@ -5,8 +5,8 @@ def flip_bbox(bbox, size, y_flip=False, x_flip=False):
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
     the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
-    where the four attributes are coordinates of the bottom left and the
-    top right vertices.
+    where the four attributes are coordinates of the top left and the
+    bottom right vertices.
 
     Args:
         bbox (~numpy.ndarray): An array whose shape is :math:`(R, 4)`.

chainercv/transforms/bbox/resize_bbox.py

@@ -5,8 +5,8 @@ def resize_bbox(bbox, in_size, out_size):
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
     the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
-    where the four attributes are coordinates of the bottom left and the
-    top right vertices.
+    where the four attributes are coordinates of the top left and the
+    bottom right vertices.
 
     Args:
         bbox (~numpy.ndarray): An array whose shape is :math:`(R, 4)`.

chainercv/transforms/bbox/translate_bbox.py

@@ -10,8 +10,8 @@ def translate_bbox(bbox, y_offset=0, x_offset=0):
     tensor of shape :math:`(R, 4)`, where :math:`R` is the number of
     bounding boxes in the image. The second axis represents attributes of
     the bounding box. They are :obj:`(y_min, x_min, y_max, x_max)`,
-    where the four attributes are coordinates of the bottom left and the
-    top right vertices.
+    where the four attributes are coordinates of the top left and the
+    bottom right vertices.
 
     Args:
         bbox (~numpy.ndarray): Bounding boxes to be transformed. The shape is

chainercv/utils/__init__.py

@@ -8,7 +8,9 @@
 from chainercv.utils.iterator import unzip  # NOQA
 from chainercv.utils.testing import assert_is_bbox  # NOQA
 from chainercv.utils.testing import assert_is_detection_dataset  # NOQA
+from chainercv.utils.testing import assert_is_detection_link  # NOQA
 from chainercv.utils.testing import assert_is_image  # NOQA
 from chainercv.utils.testing import assert_is_semantic_segmentation_dataset  # NOQA
+from chainercv.utils.testing import assert_is_semantic_segmentation_link  # NOQA
 from chainercv.utils.testing import ConstantStubLink  # NOQA
 from chainercv.utils.testing import generate_random_bbox  # NOQA

chainercv/utils/download.py

@@ -1,4 +1,6 @@
+from __future__ import division
 from __future__ import print_function
+
 import hashlib
 import os
 import shutil
@@ -13,22 +15,28 @@
 import time
 
 from chainer.dataset.download import get_dataset_directory
+from chainer.dataset.download import get_dataset_root
 
 
 def _reporthook(count, block_size, total_size):
     global start_time
     if count == 0:
         start_time = time.time()
+        print('  %   Total    Recv       Speed  Time left')
         return
     duration = time.time() - start_time
-    progress_size = int(count * block_size)
+    progress_size = count * block_size
     try:
-        speed = int(progress_size / (1024 * duration))
+        speed = progress_size / duration
     except ZeroDivisionError:
         speed = float('inf')
-    percent = int(count * block_size * 100 / total_size)
-    sys.stdout.write('\r...{}, {} MB, {} KB/s, {} seconds passed'.format(
-        percent, progress_size / (1024 * 1024), speed, duration))
+    percent = progress_size / total_size * 100
+    eta = int((total_size - progress_size) / speed)
+    sys.stdout.write(
+        '\r{:3.0f} {:4.0f}MiB {:4.0f}MiB {:6.0f}KiB/s {:4d}:{:02d}:{:02d}'
+        .format(
+            percent, total_size / (1 << 20), progress_size / (1 << 20),
+            speed / (1 << 10), eta // 60 // 60, (eta // 60) % 60, eta % 60))
     sys.stdout.flush()
 
 
@@ -51,12 +59,12 @@ def cached_download(url):
         str: Path to the downloaded file.
 
     """
-    cache_root = get_dataset_directory('_dl_cache')
+    cache_root = os.path.join(get_dataset_root(), '_dl_cache')
     try:
         os.makedirs(cache_root)
     except OSError:
         if not os.path.exists(cache_root):
-            raise RuntimeError('cannot create download cache directory')
+            raise
 
     lock_path = os.path.join(cache_root, '_dl_lock')
     urlhash = hashlib.md5(url.encode('utf-8')).hexdigest()
@@ -69,7 +77,9 @@ def cached_download(url):
     temp_root = tempfile.mkdtemp(dir=cache_root)
     try:
         temp_path = os.path.join(temp_root, 'dl')
-        print('Downloading from {}...'.format(url))
+        print('Downloading ...')
+        print('From: {:s}'.format(url))
+        print('To: {:s}'.format(cache_path))
         request.urlretrieve(url, temp_path, _reporthook)
         with filelock.FileLock(lock_path):
             shutil.move(temp_path, cache_path)

chainercv/utils/testing/__init__.py

@@ -1,6 +1,8 @@
 from chainercv.utils.testing.assertions import assert_is_bbox  # NOQA
 from chainercv.utils.testing.assertions import assert_is_detection_dataset  # NOQA
+from chainercv.utils.testing.assertions import assert_is_detection_link  # NOQA
 from chainercv.utils.testing.assertions import assert_is_image  # NOQA
 from chainercv.utils.testing.assertions import assert_is_semantic_segmentation_dataset  # NOQA
+from chainercv.utils.testing.assertions import assert_is_semantic_segmentation_link  # NOQA
 from chainercv.utils.testing.constant_stub_link import ConstantStubLink  # NOQA
 from chainercv.utils.testing.generate_random_bbox import generate_random_bbox  # NOQA

chainercv/utils/testing/assertions/__init__.py

@@ -1,4 +1,6 @@
 from chainercv.utils.testing.assertions.assert_is_bbox import assert_is_bbox  # NOQA
 from chainercv.utils.testing.assertions.assert_is_detection_dataset import assert_is_detection_dataset  # NOQA
+from chainercv.utils.testing.assertions.assert_is_detection_link import assert_is_detection_link  # NOQA
 from chainercv.utils.testing.assertions.assert_is_image import assert_is_image  # NOQA
 from chainercv.utils.testing.assertions.assert_is_semantic_segmentation_dataset import assert_is_semantic_segmentation_dataset  # NOQA
+from chainercv.utils.testing.assertions.assert_is_semantic_segmentation_link import assert_is_semantic_segmentation_link  # NOQA

chainercv/utils/testing/assertions/assert_is_detection_link.py

@@ -0,0 +1,59 @@
+import numpy as np
+import six
+
+from chainercv.utils.testing.assertions.assert_is_bbox import assert_is_bbox
+
+
+def assert_is_detection_link(link, n_fg_class):
+    """Checks if a link satisfies detection link APIs.
+
+    This function checks if a given link satisfies detection link APIs
+    or not.
+    If the link does not satifiy the APIs, this function raises an
+    :class:`AssertionError`.
+
+    Args:
+        link: A link to be checked.
+        n_fg_class (int): The number of foreground classes.
+
+    """
+
+    imgs = [
+        np.random.randint(0, 256, size=(3, 480, 640)).astype(np.float32),
+        np.random.randint(0, 256, size=(3, 480, 320)).astype(np.float32)]
+
+    result = link.predict(imgs)
+    assert len(result) == 3, \
+        'Link must return three elements: bboxes, labels and scores.'
+    bboxes, labels, scores = result
+
+    assert len(bboxes) == len(imgs), \
+        'The length of bboxes must be same as that of imgs.'
+    assert len(labels) == len(imgs), \
+        'The length of labels must be same as that of imgs.'
+    assert len(scores) == len(imgs), \
+        'The length of scores must be same as that of imgs.'
+
+    for bbox, label, score in six.moves.zip(bboxes, labels, scores):
+        assert_is_bbox(bbox)
+
+        assert isinstance(label, np.ndarray), \
+            'label must be a numpy.ndarray.'
+        assert label.dtype == np.int32, \
+            'The type of label must be numpy.int32.'
+        assert label.shape[1:] == (), \
+            'The shape of label must be (*,).'
+        assert len(label) == len(bbox), \
+            'The length of label must be same as that of bbox.'
+        if len(label) > 0:
+            assert label.min() >= 0 and label.max() < n_fg_class, \
+                'The value of label must be in [0, n_fg_class - 1].'
+
+        assert isinstance(score, np.ndarray), \
+            'score must be a numpy.ndarray.'
+        assert score.dtype == np.float32, \
+            'The type of score must be numpy.float32.'
+        assert score.shape[1:] == (), \
+            'The shape of score must be (*,).'
+        assert len(score) == len(bbox), \
+            'The length of score must be same as that of bbox.'

chainercv/utils/testing/assertions/assert_is_semantic_segmentation_link.py

@@ -0,0 +1,35 @@
+import numpy as np
+import six
+
+
+def assert_is_semantic_segmentation_link(link, n_class):
+    """Checks if a link satisfies semantic segmentation link APIs.
+
+    This function checks if a given link satisfies semantic segmentation link
+    APIs or not.
+    If the link does not satifiy the APIs, this function raises an
+    :class:`AssertionError`.
+
+    Args:
+        link: A link to be checked.
+        n_class (int): The number of classes including background.
+
+    """
+
+    imgs = [
+        np.random.randint(0, 256, size=(3, 480, 640)).astype(np.float32),
+        np.random.randint(0, 256, size=(3, 480, 320)).astype(np.float32)]
+
+    labels = link.predict(imgs)
+    assert len(labels) == len(imgs), \
+        'The length of labels must be same as that of imgs.'
+
+    for img, label in six.moves.zip(imgs, labels):
+        assert isinstance(label, np.ndarray), \
+            'label must be a numpy.ndarray.'
+        assert label.dtype == np.int32, \
+            'The type of label must be numpy.int32.'
+        assert label.shape == img.shape[1:], \
+            'The shape of label must be (H, W).'
+        assert label.min() >= 0 and label.max() < n_class, \
+            'The value of label must be in [0, n_class - 1].'

docs/source/reference/utils.rst

@@ -57,19 +57,27 @@ Testing Utilities
 
 assert_is_bbox
 ~~~~~~~~~~~~~~
-.. autofunctions:: assert_is_bbox
+.. autofunction:: assert_is_bbox
 
 assert_is_detection_dataset
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~
-.. autofunctions:: assert_is_detection_dataset
+.. autofunction:: assert_is_detection_dataset
+
+assert_is_detection_link
+~~~~~~~~~~~~~~~~~~~~~~~~
+.. autofunction:: assert_is_detection_link
 
 assert_is_image
 ~~~~~~~~~~~~~~~
-.. autofunctions:: assert_is_image
+.. autofunction:: assert_is_image
 
 assert_is_semantic_segmentation_dataset
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-.. autofunctions:: assert_is_semantic_segmentation_dataset
+.. autofunction:: assert_is_semantic_segmentation_dataset
+
+assert_is_semantic_segmentation_link
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+.. autofunction:: assert_is_semantic_segmentation_link
 
 ConstantStubLink
 ~~~~~~~~~~~~~~~~

tests/links_tests/model_tests/faster_rcnn_tests/test_faster_rcnn.py

@@ -5,6 +5,8 @@
 from chainer import testing
 from chainer.testing import attr
 
+from chainercv.utils import assert_is_detection_link
+
 from dummy_faster_rcnn import DummyFasterRCNN
 
 
@@ -58,39 +60,13 @@ def test_call_gpu(self):
         self.link.to_gpu()
         self.check_call()
 
-    def check_predict(self):
-        imgs = [
-            _random_array(np, (3, 640, 480)),
-            _random_array(np, (3, 320, 320))]
-
-        bboxes, labels, scores = self.link.predict(imgs)
-
-        self.assertEqual(len(bboxes), len(imgs))
-        self.assertEqual(len(labels), len(imgs))
-        self.assertEqual(len(scores), len(imgs))
-
-        for bbox, label, score in zip(bboxes, labels, scores):
-            self.assertIsInstance(bbox, np.ndarray)
-            self.assertEqual(bbox.dtype, np.float32)
-            self.assertEqual(bbox.ndim, 2)
-            self.assertLessEqual(bbox.shape[0], self.n_roi)
-            self.assertEqual(bbox.shape[1], 4)
-
-            self.assertIsInstance(label, np.ndarray)
-            self.assertEqual(label.dtype, np.int32)
-            self.assertEqual(label.shape, (bbox.shape[0],))
-
-            self.assertIsInstance(score, np.ndarray)
-            self.assertEqual(score.dtype, np.float32)
-            self.assertEqual(score.shape, (bbox.shape[0],))
-
     def test_predict_cpu(self):
-        self.check_predict()
+        assert_is_detection_link(self.link, self.n_class - 1)
 
     @attr.gpu
     def test_predict_gpu(self):
         self.link.to_gpu()
-        self.check_predict()
+        assert_is_detection_link(self.link, self.n_class - 1)
 
 
 @testing.parameterize(