Alejandro Newell

If coordinates are outside of the image frame, the ground truth heatmap consists of all zeros, the same loss (mean squared error) is applied.

The training is different for the three settings, the network is either provided ground truth detections as input (in PR for example) or needs to produce detections from scratch (for...

https://github.com/umich-vl/pose-ae-train/issues/5

There are utilities to draw networks that use nngraph, you can see some examples in the [nngraph readme](https://github.com/torch/nngraph/blob/master/README.md). Though I personally haven't used it so not sure how the visualization...

Hi @mkocabas This is a tough issue, especially on the COCO data. The way that normalization is done during the official COCO evaluation is by pixel area of the person's...

Sorry for the confusion about extra output dimensions, we only use 34. The additional output channels are leftover from an old idea but are not relevant to the current model.

Hi, I don't have the file written for flic.lua, and probably won't end up writing one, sorry! We are working on cleaning up a code release in TensorFlow for the...

I don't think it is included here, but you can find the code for it in the training repo here: https://github.com/umich-vl/pose-hg-train/blob/master/src/util/pose.lua#L85

I've seen that error with either an outdated version of Torch or the pretrained model (don't remember which), so one or the other likely needs an update. If that doesn't...

I just mentioned that in case you happened to have downloaded the model months ago, which I don't think is the issue. I still think the likely culprit is an...