How do I load images from URLs into tensorflow?

Question

I'm currently running this:

image = urllib.urlopen(imgUrl)
pool3_features = sess.run(pool3,{'incept/DecodeJpeg/contents:0': image})

and I get this error:

Unable to get element from the feed as bytes.

Can you post a full runable code, What is the meaning of the `gfile`? — bwangel, Feb 14 '17 at 07:33
sorry, I want to place the img data into a [bytesio](https://docs.python.org/3/library/io.html#binary-i-o), but i didn't find any api can be used in tensorflow. — bwangel, Feb 15 '17 at 05:58

score 3 · Accepted Answer · answered Feb 21 '17 at 06:52

3

The solution was quite simple... All I had to do was call the read method on response from urlopen. The following works like a charm:

image = urllib.urlopen(imgUrl)
pool3_features = sess.run(pool3,{'incept/DecodeJpeg/contents:0': image.read()})

answered Feb 21 '17 at 06:52

JJJ

score -1 · Answer 2 · answered Feb 15 '17 at 19:37

-1

Just use a method to load it into NumPy and then load that into TensorFlow. This is likely more reliable, more flexible.

answered Feb 15 '17 at 19:37

aselle

2 Answers2