Ronghang Hu

There isn't a difference in principle. The summation operation is used for output in Compact Bilinear Pooling (http://arxiv.org/abs/1511.06062, see their Figure 1), so I also added this option.

You can explicitly set the (static) output shapes using [`Tensor.set_shape`](https://www.tensorflow.org/api_docs/python/tf/Tensor#set_shape).

Hi, we used a Jupyter notebook to visualize the trajectories. Here is the notebook we used: https://gist.github.com/ronghanghu/d250f3a997135c667b114674fc12edae

For that part, we don't have an automated script to generate them (but we loaded them in web browser and screenshot them). This [3D.zip](https://github.com/ronghanghu/speaker_follower/files/2522092/3D.zip) is a javascript webGL script from...

Hi, to run the 3D visualizer, one would need to: * download the meshgrid data of Matterport3D * add the .js extension to files in the js directory * set...

This is unexpected, and is probably due to the change in tensorflow versions (the code is written with TF 0.7.0 and is probably not compatible with newer version of tensorflow)....

Mmm, that's very strange. Probably the training batches weren't correctly built? Blow is my terminal output when re-training the model using this repo. The output when training the high resolution...

I'm not exactly sure what's wrong. The final accuracy `0.238127 (all), 0.418813 (pos), 0.216635 (neg)` you got seems even lower than random guess (0.50 on this binary classification problem) and...

Btw, if you are in a hurry and the problem persists, you can try out the Caffe implementation of this project (https://github.com/Seth-Park/text_objseg_caffe implemented by Seth Park), which gives higher performance...

Hi, sorry that the links have expired. I've uploaded the trained models and the augmented data here: https://drive.google.com/drive/folders/1yCzG3g9vPlatFWokJbxZp8nWuN-xCNLW Hope it helps!