in

Tensorboard visualization with Pointnet2 and Google colab



We train Pointnet2 on Google colab and visualize the accuracy and loss using Tensorboard.

and our forked repository:

https://github.com/mhwasil/pointnet2

To visualize the accuracy and loss on Google colab, please refer to this tutorial:
https://stackoverflow.com/questions/47818822/can-i-use-tensorboard-with-google-colab

Please note that we generated the customized tensor operators (under tf_ops) locally using tensorflow 1.4.1 and cuda-8.0, and upload the generated tensor. After that, pull the generated tensor to Google colab and reinstall tensorflow on it to 1.4.1.

All the generated tensor operators were uploaded to our forked pointnet2 under “colab” branch.

Share this: