How is the dataset divided?

[xushenlz]

How is the dataset divided?

[XuShenLZ]

Thanks for your interest! I don't think we have received the dataset requests from you. May I know which data files did you use? Also, could you please describe your experiment steps clearly?

[yuwenyouhaoran]

Thanks for your interest! I don't think we have received the dataset requests from you. May I know which data files did you use? Also, could you please describe your experiment steps clearly?

I'm sorry that the previous description was not clear.

I just cloned the code and used the default parameters and DJI_0007~DJI_0014 in the code for training, but the results obtained are different from those in the paper. May I ask which part of the data was used for training to get the results in the paper?

The dataset was applied by my classmates in the same lab, and I'm doing similar research with him, therefore using the same dataset.

[XuShenLZ]

The default code is only using DJI_0006 for fast execution. You definitely need more training data. In the paper, we are using DJI_0007~0022 which results in roughly 57500 total data points. Of those, about 51750 were used for training, and the rest for test + validation.

So you might want to change the code to be ["DJI_" + str(i).zfill(4) for i in range(7, 23)] in both create_dataset.py and prepar_single_file_data.py. You might find the number becoming a little bit different now when you run the code since we slightly changed a few data processing after paper submission. But the result shouldn't deviate too much.

In terms of the dataset, please limit the circulation inside your lab. Please let others submit another request if they are interested. Thanks!

[yuwenyouhaoran]

Thank you very much for your help in your busy schedule! I have fully understood the process of your experiment through your detailed explanation.