Jonghoon Jin
Jonghoon Jin
I tried to make this Makefile https://github.com/jhjin/tensorflow-cpp/blob/master/Makefile self-explanatory. `make app` and `make run` will tell you how to compile the example app and execute it. Note that this script is...
Hi, this is a handy example to show how to build / use TF C++ interface. I do not have a counter part training / freezing code.
You need to build tensorflow from the source but this will be automatically done by triggering make target. The first thing to do is to install dependency with `make dependency`....
Hi I haven't checked the configuration recently. My guess is that your Bazel version is newer than the one I used to set up this repository. Either you can switch...
It should be doable as long as you have right dependencies installed. The Makefile has no information about fedora dependencies, feel free to make a PR. On Mon, Mar 20,...
One possible reason I thought was the difference in input preprocessing. The network model requires a fixed size of input, which needs to be converted from raw image. I think...
If the negative anchor is not found, loss is calculated according to the worst case. Which is reasonable most of time as long as batch size is big enough.
Thanks for sharing @soumith I learn a lot from your work