【Kinetics Dataset】Data Corrupted or incomplete from Baiduyun or Google Drive about mmskeleton HOT 5 CLOSED

@czqInNanjing
Hi, both documents in Baiduyun and Google Drive are correct, and the script for building database works well on my server.
I am wondering how you unarchive the zip files on Baidu Yun.
Since users can not directly download large documents in BaiduYun from the browser. I split the zip document into 2 segments. To extract files from multiple zip segments, you should:

1. collect 2 parts together.
2. run zip -F kinetics-skeleton.zip --out merge.zip
3. run unzip ./merge.zip

from mmskeleton.

Hi! Thank you for reporting this problem. I will check it as soon as possible.

from mmskeleton.

Thank you for your kind answer!

I unarchive the zip both in Mac and in Ubuntu 16.04.
In mac, using the "Unarchive"
In ubuntu, using the default "Archive Manager" to unarchive.
Under both situation, it reports file is corrupted.

But, we still continue to build with these files.
Although the script for building dataset still failed on my PC and my collaborator's PC, we found that it only fail for 4 samples by change the code

instead of "assert", we print the sample that does not match
assert (self.label[index] == label) feeder_kinetics.py Line 131
===>
print(self.label[index])

so we jump through these samples and successfully build the dataset.

The code now work well.

from mmskeleton.

The Baiduyun link is out of date, can you provide a new one? I can not download file from Google Drive.

from mmskeleton.

The Baiduyun link is out of date, can you provide a new one? I can not download file from Google Drive.

请问有百度云吗

from mmskeleton.