Swathikiran Sudhakaran

Have you succeeded in loading the caffe2 weights to the provided pytorch model?

You can apply the hidden state of the first CLSTM cell as input to the second CLSTM cell. Please see the code below: ``` def __init__(self, mem_size): super(deepSignal, self).__init__() self.mem_size...

self.convNet = nn.Sequential(*list(self.resnet.children())[:-2]) Change the input dimension of ConvLSTM to 512 for ResNet18 or 2048 for ResNet50 in [here](https://github.com/swathikirans/violence-recognition-pytorch/blob/master/createModel.py#L12)

I don't think it is because of the change in the version of pytorch or python. If you are loading the images from a harddisk it will take a lot...

The second case indeed makes more sense. However, I am not sure if it would make a significant impact on the final performance of the model. I will update the...

Hi, temporal information is injected by shifting and the gating select how much of the features are to be shifted. The code do not have the edge cases of gating,...

Hi, there is no predefined rule for feature selection. The module learns from the input data. In short, the routing of features is data dependent and not hardwired.

We trained the network with action level supervision and during inference separated the action to verb-noun pairs. This was done to analyze how the network improved on verb and noun....

The activity labels are of the format "verb_nounGroup". I mapped the activity labels (from the output of the network) to the names and separated them after the first '_' to...

@KrisLee512 and @i-amgeek, indeed you need to download the data beforehand. You may check this repo for information regarding GTEA61 dataset setup https://github.com/swathikirans/ego-rnn