Closed taojake closed 6 years ago
I'm not sure I can completely follow your question. We don't to any object tracking. We also don't provide code for freezing the model so you will have to write it yourself. If that code doesn't work it will be hard to offer support here. I know from experience that you can get it to work though. You will have to feed the input and fetch the output from the endpoints['emb']
tensor.
I added a freeze.py
file that freezes the model, here: https://github.com/arcelien/triplet-reid. It's probably easiest to add a name to endpoints['emb']
to reference it once you freeze it.
I have figured out it. Thanks a lot. @arcelien
FYI: I just merged @arcelien's freeze code.
i want to use pre-trained model in deploy, so, i want to transfer checkpoint to pb file, but when i use emb/biases node, there only freeze this node to pb, so, which node is the output for object tracking inference