kinhong/OpenLabeler

Any plans to support Tensorflow 2?

veonua opened this issue · 4 comments

Object detection API has been updated to run on the latest TF version,
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md

it seems like output vector has been replaced with a dictionary, at least in my code I had to replace

num_detections = result[0].numpy()
classes = result[1].numpy()

to

num_detections = result["num_detections"].numpy()
classes = result['detection_classes'].numpy()

@veonua I am still waiting for a more official release of TensorFlow 2.0 Java API (I have been following this issue)

References:
https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow-2.3.0.jar
https://storage.googleapis.com/tensorflow/libtensorflow/libtensorflow_jni-cpu-linux-x86_64-2.3.0.tar.gz
https://www.tensorflow.org/install/lang_java

looks promising, but I am not fully aware of set of jars required to run it on Java, I assume update of protobuf would be good enough to run inference.

@veonua2 In release v1.3.0, OpenLabeler now supports TensorFlow 2.3.1.

thank you very much