evaluate on test dataset when training?

X-Temporal/x_temporal/interface/temporal_helper.py

Line 386 in 5faa5f4

test_loader = self.data_loaders['test']

Here the data_loaders is set to "test" at any time.
I'm not sure wether this is a bug or not. Maybe there is something I misunderstand?

It may be a bug, I will update the code later. You can easily modify the code if you need to change it now.

maybe we should set test_mode=test_mode here:
https://github.com/Sense-X/X-Temporal/blob/master/x_temporal/utils/dataset_helper.py#L27

The arg test_mode in the lastest version didn't affect the training or testing, we will update the code to fix this problem.
Thanks for your suggestions

Is there any difference between _get_val_indices() and _get_test_indices()?
https://github.com/Sense-X/X-Temporal/blob/master/x_temporal/core/dataset.py#L173

In this version, they have no difference.

https://github.com/Sense-X/X-Temporal/blob/master/x_temporal/core/dataset.py#L220
https://github.com/Sense-X/X-Temporal/blob/master/x_temporal/core/dataset.py#L313

Here we count image frames from 1 and count video from 0. This difference would lead to "index out of bounds" error.
We should change "seg_ind" to "seg_ind - 1" for video_source in https://github.com/Sense-X/X-Temporal/blob/master/x_temporal/core/dataset.py#L313.

Thanks for your suggestion, I have fixed this bug.