Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

consider dilation in cnn_output_size Calculation of python api #5245

Closed
wants to merge 2 commits into from
Closed

consider dilation in cnn_output_size Calculation of python api #5245

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
021e1fa
fix outputsize bug with dilation
NHZlX Oct 31, 2017
bcd6a71
M:erge branch 'develop' of https://github.com/PaddlePaddle/Paddle int…
NHZlX Oct 31, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
30 changes: 22 additions & 8 deletions python/paddle/trainer/config_parser.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1200,8 +1200,14 @@ def TestData(data_config, async_load_data=None):

#caffe_mode: compute the output size using floor instead of ceil,
# which is consistent of caffe and CuDNN's convention.
def cnn_output_size(img_size, filter_size, padding, stride, caffe_mode):
output = (2 * padding + img_size - filter_size) / float(stride)
def cnn_output_size(img_size,
filter_size,
padding,
stride,
caffe_mode,
dilation=1):
filter_s = (filter_size - 1) * dilation + 1
output = (2 * padding + img_size - filter_s) / float(stride)
if caffe_mode:
return 1 + int(math.floor(output))
else:
Expand All @@ -1210,8 +1216,14 @@ def cnn_output_size(img_size, filter_size, padding, stride, caffe_mode):

#calcualte image_size based on output_size for de-convolution (ConvTransLayer).
#It is the reverse function of cnn_output_size
def cnn_image_size(output_size, filter_size, padding, stride, caffe_mode):
img_size = (output_size - 1) * stride + filter_size - 2 * padding
def cnn_image_size(output_size,
filter_size,
padding,
stride,
caffe_mode,
dilation=1):
filter_s = (filter_size - 1) * dilation + 1
img_size = (output_size - 1) * stride + filter_s - 2 * padding
if not caffe_mode:
img_size = img_size + 1
return img_size
Expand Down Expand Up @@ -1376,27 +1388,29 @@ def parse_conv(conv, input_layer_name, conv_conf, num_filters, trans=False):
conv_conf.stride_y = conv.stride_y
conv_conf.groups = conv.groups
conv_conf.caffe_mode = conv.caffe_mode
conv_conf.dilation = conv.dilation
conv_conf.dilation_y = conv.dilation_y

if not trans:
conv_conf.filter_channels = conv.channels / conv.groups
conv_conf.img_size, conv_conf.img_size_y = \
get_img_size(input_layer_name, conv.channels)
conv_conf.output_x = cnn_output_size(
conv_conf.img_size, conv_conf.filter_size, conv_conf.padding,
conv_conf.stride, conv_conf.caffe_mode)
conv_conf.stride, conv_conf.caffe_mode, conv_conf.dilation)
conv_conf.output_y = cnn_output_size(
conv_conf.img_size_y, conv_conf.filter_size_y, conv_conf.padding_y,
conv_conf.stride_y, conv_conf.caffe_mode)
conv_conf.stride_y, conv_conf.caffe_mode, conv_conf.dilation_y)
else:
conv_conf.filter_channels = num_filters / conv.groups
conv_conf.output_x, conv_conf.output_y = \
get_img_size(input_layer_name, conv.channels)
conv_conf.img_size = cnn_image_size(
conv_conf.output_x, conv_conf.filter_size, conv_conf.padding,
conv_conf.stride, conv_conf.caffe_mode)
conv_conf.stride, conv_conf.caffe_mode, conv_conf.dilation)
conv_conf.img_size_y = cnn_image_size(
conv_conf.output_y, conv_conf.filter_size_y, conv_conf.padding_y,
conv_conf.stride_y, conv_conf.caffe_mode)
conv_conf.stride_y, conv_conf.caffe_mode, conv_conf.dilation_y)


#caffe_mode: compute the output size using floor instead of ceil,
Expand Down