Sounds overlap between test set and train set

[amiryanj]

The train and set set has completely overlap

In file create_dataset.py, lines 118-120 the frame indices are shuffled and form the train_idx and test_idx.
Let's assume frames 10 is in test_idx and frame 9 is in train_idx.
Then when you call get_samples_from_frame(f_i=9), it would use the information of frame 10 to generate the samples (lines 81:82).
Then let's say the test data almost perfectly exist in train_set as well!!!
Is there something I misunderstood?

[piaozhx]

Actually, you are right, It's an old version of the function create_GC_train_test_data, which is a mistake when we release create_dataset.py.

an easy solution is:

# wrong! it will cause overlap  between test set and train set
# random_idx = random.sample(range(frame_num), frame_num)

random_idx = range(frame_num)

[amiryanj]

Yes I fixed that mistake but it seems that results will converge somewhere else.
I run the algorithm for 350 epochs (10,000 is too much!) and the error does not come under 0.017.