Gunnar Sigurdsson

Strange, maybe these posts have some information: https://github.com/gsig/PyVideoResearch/issues/13 https://github.com/gsig/PyVideoResearch/issues/11 There is a tiny chance that there has been some code drift since the baseline that would cause these discrepancies, which...

Ah, good catch. The readme was misleading, updated now. aj_rgb_imagenet.pth is a model trained on imagenet+kinetics aj_rgb_charades.pth is the aj_rgb_imagenet.pth model further trained on Charades. If you need a model...

It's been while since I experimented with the i3d-inception models on Charades, since I mostly switched to ResNet50-I3D, except for trying to replicate the AVA baseline. One thing to keep...

I have not experimented with ResNet101-3D much, so you might indeed have to experiment with the hyperparameters and double check the 101 extension, but it should work similarly to the...

You mean for evaluation? The bounding boxes (that have arbitrary dimensions) are loaded under the `meta` field: https://github.com/gsig/PyVideoResearch/blob/46307b1a03ce670696297e2154ddee6f4e6b0b8a/models/criteria/maskrcnn_criterion.py#L46 This is then returned as `score_targets` which is used for evaluation by...

Let us know if you figured it out, and write up your experiences to help others if you have time!

The error is complaining that a passed keyword argument to TSN base is somehow a tuple, not a dict. I have not tested the TSN baseline recently, so there is...

The TSN part of the codebase was very experimental, so feel free to submit a pull request if you get good results. I believe this is a version of TSN...

Currently, Tensorflow is unfortunately not supported. There might be third-party code lying around that uses tensorflow to load charades data, such as https://github.com/rohitgirdhar/ActionVLAD

The code for the message passing is here: [https://github.com/gsig/temporal-fields/blob/master/pytorch/models/layers/AsyncTFCriterion.py](https://github.com/gsig/temporal-fields/blob/master/pytorch/models/layers/AsyncTFCriterion.py) The rest of the code is just a deep network, the predicts the potentials for this graphical model. hope that helps,...