Read images from urls

eta/core/image.py

@@ -391,13 +391,13 @@ def download(url, include_alpha=False, flag=None):
     return decode(bytes, include_alpha=include_alpha, flag=flag)
 
 
-def read(path, include_alpha=False, flag=None):
-    '''Reads image from path.
+def read(path_or_url, include_alpha=False, flag=None):
+    '''Reads image from a file path or url.
 
     By default, images are returned as color images with no alpha channel.
 
     Args:
-        path: the path to the image on disk
+        path_or_url: the file path or url to the image
         include_alpha: whether to include the alpha channel of the image, if
             present, in the returned array. By default, this is False
         flag: an optional OpenCV image format flag to use. If provided, this
@@ -406,10 +406,13 @@ def read(path, include_alpha=False, flag=None):
     Returns:
         a uint8 numpy array containing the image
     '''
+    if etaw.is_url(path_or_url):
+        return download(path_or_url, include_alpha=include_alpha, flag=flag)
+
     flag = _get_opencv_imread_flag(flag, include_alpha)
-    img_bgr = cv2.imread(path, flag)
+    img_bgr = cv2.imread(path_or_url, flag)
     if img_bgr is None:
-        raise OSError("Image not found '%s'" % path)
+        raise OSError("Image not found '%s'" % path_or_url)
     return _exchange_rb(img_bgr)
 
 

eta/core/web.py

@@ -20,7 +20,9 @@
 # pragma pylint: enable=wildcard-import
 
 import logging
+import re
 import requests
+import six
 from time import time
 
 import eta.constants as etac
@@ -30,6 +32,22 @@
 logger = logging.getLogger(__name__)
 
 
+URL_REGEX = re.compile(r'http://|https://|ftp://|file://|file:\\')
+
+
+def is_url(filename):
+    '''Return True if string is an http or ftp path.
+
+    Args:
+        filename: a string
+
+    Returns:
+        True/False if the filename is a url
+    '''
+    return (isinstance(filename, six.string_types) and
+            URL_REGEX.match(filename) is not None)
+
+
 def download_file(url, path=None, chunk_size=None):
     '''Downloads a file from a URL. If a path is specified, the file is written
     there. Otherwise, the content is returned as a binary string.