1 | initial version |
@eithne,
I tried fiddling with all the parameters passed to blobFromImage, but nothing I tried made it as good as using TensorFlow's API.
There is a table with one and only preprocessing parameters: scale factor 0.00784
, mean subtraction 127.5 127.5 127.5
and resizing to 300x300.
Have you tried to use OpenCV's sample applications? object_detection.cpp, object_detection.py