Applying Transforms

[LukeDitria]

Hi,
Thanks for the dataset.
I noticed that when applying a transform the sequence seems to reverse, this appears to be due to that fact that within the _transform_time method you are concatenating the current image to the start of the sequence instead of the end, also the input sequence transform is applied to both the sequence and target and the target transform is never used, I believe the function should look something like:

def _transform_time(data, target=False):
  new_data = None
  for i in range(data.size(0)):
      img = Image.fromarray(data[i].numpy(), mode='L')
      if target:
          new_data = self.target_transform(img) if new_data is None else torch.cat([new_data, self.target_transform(img)], dim=0)
      else:
          new_data = self.transform(img) if new_data is None else torch.cat([new_data, self.transform(img)], dim=0)
  return new_data

[dnl-trpp]

Hi! Nice project. Just faced the same problem and indeed the sequence reverses when using transforms. Would be nice to see a pull request on this :)

[tychovdo]

Thanks. I just put the dataset on public after using it for a toy project in case anyone might find it useful.
It's not tested and I won't be maintaining it, but feel free to send in a pull request then I'll merge it in.