Trying to load a model. Need to update JSON parsing and the layer interface to support batching.
Second Build: 1.13 with Training
Unsure if this is possible yet but see the org.tensorflow.* package. It looks like it contains all the objects needed for this kind of thing. But we may not have full op support for the classification models we test against and almost certainly not for our own models.
Third Build: 1.15 | 2.0 with Inference/Training
Custom build from our doc.ai fork.
2.0 does have a build_all_android script in tensorflow/contrib/makefile and there are other instructions around the net for building tensorflow via bazel. JNI and NDK are required.
First Build: 1.13 with Inference
As far as I can tell, last build officially supplied by the TensorFlow team. Use:
Trying to load a model. Need to update JSON parsing and the layer interface to support batching.
Second Build: 1.13 with Training
Unsure if this is possible yet but see the
org.tensorflow.*
package. It looks like it contains all the objects needed for this kind of thing. But we may not have full op support for the classification models we test against and almost certainly not for our own models.Third Build: 1.15 | 2.0 with Inference/Training
Custom build from our doc.ai fork.
2.0 does have a build_all_android script in tensorflow/contrib/makefile and there are other instructions around the net for building tensorflow via bazel. JNI and NDK are required.