The text recognition 0012 model (http://docs.openvinotoolkit.org/2019_R1/_text_recognition_0012_description_text_recognition_0012.html) has the following input and output:
[B x C x H x W] -> [ 1 x 1 x 32 x 120]
B - batch size
C - number of channels
H - image height
W - image width
[W x B x L] -> [30 x 1 x 37]
W - output sequence length
L - confidence distribution across alpha-numeric symbols: "0123456789abcdefghijklmnopqrstuvwxyz#", where # - special blank character for CTC decoding algorithm.
So, I need to change the network batch size. To do that I used the following method:
After using the above method the network output was:
Batch_size = 2 Output = [60 x 1 x 37]
Batch_size = 3 Output = [90 x 1 x 37]
The correct output, for me, should be:
Batch_size = 2 Output = [60 x 2 x 37]
Batch_size = 3 Output = [90 x 3 x 37]
Am I wrong? How can I change the network batch size correctly?
For more complete information about compiler optimizations, see our Optimization Notice.