Andrew Hundt

@tnikolla can you explain the constants 0.35 and 0.47? They appear all over the place, such as in bboxes to grasps, grasp to bbox, and in the iou calculation.

This repository suffers from averaging all data, unfortunately. For example if there is a frisbee it will try to grab the center rather than the lid edge. I've got improved...

sorry, I haven't done kalibr setup in a long time. If you fix something please make a pull request! It would be very appreciated

@mtourne can you incorporate the new training scripts from master? @bittnt any chance of merging this?

@mtourne I tried building your branch and got quite a few compiler errors [build_log.txt](https://github.com/torrvision/crfasrnn/files/211461/build_log.txt) I'm on ubuntu 14.04, (there were linker errors but those would be easy to resolve after...

I'd have it be a "private" (leading underscore) member function that can be called from the constructor, but otherwise that sounds good to me.

Note I've got an outstanding feature request for the necessary operations on tensorflow itself at https://github.com/tensorflow/tensorflow/issues/12948.

Cool thanks! I just wanted to confirm, sorry I missed those lines.

I wonder, should the data be put in sparse tensors? I would probably need to test to see if it is a faster or slower approach.

Fair enough, I'm actually looking both at this model and to train others on coco. It looks like the upstream repository you link doesn't train on coco, so hopefully that...