Extracting human36 parsed images

[vigneshrk29]

Hello,

I am trying to extract the parsed images of human36m. I could only get images.tar.gzaa to work the sh command. But I get the following error: gzip: stdin: unexpected end of file images/s_05_act_06_subact_01_ca_02/s_05_act_06_subact_01_ca_02_000871.jpg tar: Unexpected EOF in archive tar: Unexpected EOF in archive tar: Error is not recoverable: exiting now

Which results in the following error when I run test.py: OSError: Fail to read ../data/Human36M/images/s_09_act_02_subact_01_ca_01/s_09_act_02_subact_01_ca_01_000001.jpg

How do I fix this?

[vigneshrk29]

Additionally, do I need all the images.tar.gzaa files to run test.py?

[mks0601]

Download all tar.gz* and run sh extract_targz.sh

[vigneshrk29]

Hi,

I downloaded all the files and ran all the data. Still some folders required for testing are missing. Like s_11_act_06_subact_02_ca_02. Is it possible to just share the parsed data of subjects 9 and 11? Else how to proceed?

Thanks

[mks0601]

I think you didn't download properly.

1. Download all tar.gz* and do not change their name
2. check md5sum
3. run the script that I uploaded (extract_targz.sh)

[vigneshrk29]

Hi,

I re-downloaded everything, ran md5sum and ran the script. Still the error that some images are missing keeps occuring. Is there: 1 ) a way to specify a specific folder from the downloaded images to run testing 2) Just get the parsed data for subject 11 and 9

[mks0601]

I've just downloaded and checked there is no problem. I can see s_11_act_06_subact_02_ca_02. Are you using Window? To answer your questions, there is no way for that.

[vigneshrk29]

Yes, I am working on windows (WSL). The data is unzipped to an external hardisk as I dont have that much space on my laptop. For example, how many times I run the extract code, the folder s09_act_02_subact_01_01 is never unzipped

[mks0601]

I guess that is the problem. I have seen some guys fail to unzip data with Window. Do you have any linux machines?

[vigneshrk29]

I have windows subsystem for linux but the space is not sufficient. Is there anyway to make it work on a windows system?

[mks0601]

sorry I have no window machine, so I can't do any test..

[tribeband]

I run the command in Linux. but it still return wrong. File "E:\Tools\anaconda\envs\cv\lib\site-packages\torch\utils\data_utils\collate.py", line 163, in collate_tensor_fn return torch.stack(batch, 0, out=out) RuntimeError: stack expects each tensor to be equal size, but got [15, 3] at entry 0 and [16, 3] at entry 1

The files download saved in the harddisk: augmented_set.tar-004.gzaa, augmented_set.tar-007.gzab, augmented_set.tar-005.gzac, augmented_set.tar-006.gzad

[mks0601]

Please do not change the filenames. The original ones do not have -004, -007 like this. And run extract_targz.sh to unzip them