juliusshufan commented on a change in pull request #9305: Fixing two symbol
files for CNN
URL: https://github.com/apache/incubator-mxnet/pull/9305#discussion_r159794950
##########
File path: example/image-classification/symbols/googlenet.py
##########
@@ -65,7 +65,7 @@ def get_symbol(num_classes = 1000, **kwargs):
pool5 = mx.sym.Pooling(in4e, kernel=(3, 3), stride=(2, 2), pool_type="max")
in5a = InceptionFactory(pool5, 256, 160, 320, 32, 128, "max", 128,
name="in5a")
in5b = InceptionFactory(in5a, 384, 192, 384, 48, 128, "max", 128,
name="in5b")
- pool6 = mx.sym.Pooling(in5b, kernel=(7, 7), stride=(1,1), pool_type="avg")
+ pool6 = mx.sym.Pooling(in5b, kernel=(7, 7), stride=(1,1), pad=(1, 1),
pool_type="avg")
Review comment:
Hi,
Thanks for the reply, do you mean I should padding the initial input shape
when using this symbol file (googlenet.py)? Actually when I using the
"benchmark_score.py" under "example/image-classification/", and the input shape
is 3x224x224, and this should accord with the initial parper introducing the
GoogleNet. So far I am avoiding the issue by adding a pad as the PR show, may I
have your comments?
Thanks.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
