A Tensorflow model for text recognition (CNN + seq2seq with visual attention) available as a Python package and compatible with Google Cloud ML Engine.
I try to use the exported model in a Java environment. I have other network models working well. Aocr is working pretty well from CLI as well. But i can not get to run in the Java library (1.15). The big difference to other models is the input tensor. It is named 'input_image_as_bytes' in aocr. The question is what type of input is expected here. I tried the raw JPG but it raises errors then:
java.lang.IllegalStateException: Attempting to use uninitialized value conv_conv5/BatchNorm/gamma
[[{{node conv_conv5/BatchNorm/gamma/read}}]]
at org.tensorflow.Session.run(Native Method) ~[libtensorflow-1.15.0.jar:1.15.0]
at org.tensorflow.Session.access$100(Session.java:48) ~[libtensorflow-1.15.0.jar:1.15.0]
at org.tensorflow.Session$Runner.runHelper(Session.java:326) ~[libtensorflow-1.15.0.jar:1.15.0]
at org.tensorflow.Session$Runner.run(Session.java:276) ~[libtensorflow-1.15.0.jar:1.15.0]
I try to use the exported model in a Java environment. I have other network models working well. Aocr is working pretty well from CLI as well. But i can not get to run in the Java library (1.15). The big difference to other models is the input tensor. It is named 'input_image_as_bytes' in aocr. The question is what type of input is expected here. I tried the raw JPG but it raises errors then: