[FFmpeg-cvslog] convert_from_tensorflow.py: add support when kernel size is 1*1 with one input/output channel (gray image)

Guo, Yejun git at videolan.org
Fri Dec 13 16:45:13 EET 2019


ffmpeg | branch: master | Guo, Yejun <yejun.guo at intel.com> | Fri Nov 22 15:50:04 2019 +0800| [e52070e89c755bffe08ecb9d9ea6be73ca1509cd] | committer: Pedro Arthur

convert_from_tensorflow.py: add support when kernel size is 1*1 with one input/output channel (gray image)

Signed-off-by: Guo, Yejun <yejun.guo at intel.com>
Signed-off-by: Pedro Arthur <bygrandao at gmail.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=e52070e89c755bffe08ecb9d9ea6be73ca1509cd
---

 tools/python/convert_from_tensorflow.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tools/python/convert_from_tensorflow.py b/tools/python/convert_from_tensorflow.py
index 605158a32e..5e87e227ea 100644
--- a/tools/python/convert_from_tensorflow.py
+++ b/tools/python/convert_from_tensorflow.py
@@ -193,7 +193,10 @@ class TFConverter:
         filter_width = ktensor.tensor_shape.dim[1].size
         in_channels = ktensor.tensor_shape.dim[2].size
         out_channels = ktensor.tensor_shape.dim[3].size
-        kernel = np.frombuffer(ktensor.tensor_content, dtype=np.float32)
+        if filter_height * filter_width * in_channels * out_channels == 1:
+            kernel = np.float32(ktensor.float_val[0])
+        else:
+            kernel = np.frombuffer(ktensor.tensor_content, dtype=np.float32)
         kernel = kernel.reshape(filter_height, filter_width, in_channels, out_channels)
         kernel = np.transpose(kernel, [3, 0, 1, 2])
 



More information about the ffmpeg-cvslog mailing list