search

moberweger / deep-prior-pp

Improving Fast And Accurate 3D Hand Pose Estimation
GNU General Public License v3.0
144 stars 41 forks source link

use ICVLImporter and NYUImporter #32

Open zhanglong321 opened 5 years ago

zhanglong321 commented 5 years ago

Hi! Thanks for you code. I'm intersted in hand pose estimation, I try to use the ICVLImporter and NYUImporter " di = ICVLImporter(basepath) Seq0 = di.loadSequence('train', ['0'], shuffle=True, rng=rng) Seq1 = di.loadSequence('test_seq_1')" but I get the error that"FileNotFoundError: [Errno 2] No such file or directory: 'F:/Hand Pose Dataset/ICVL Hand Posture Dataset/Training/train.txt'" I found that there have no such label file in the icvl dataset, just a "labels.txt", it include about 330K lines. And the NYUImporter also has error. How can I solve it? Can you tell me some details? Thank you for you helping!

moberweger commented 5 years ago

The ICVLImporter was designed for the "old" ICVL dataset, which was released 2014. You might need to adjust the file formats to fit to the new dataset. For the NYU issues, you might take a look at issue #4

zhanglong321 commented 5 years ago

The ICVLImporter was designed for the "old" ICVL dataset, which was released 2014. You might need to adjust the file formats to fit to the new dataset. For the NYU issues, you might take a look at issue #4

Thank you very much. I will try to do it. Have a good day!

zhanglong321 commented 5 years ago

The ICVLImporter was designed for the "old" ICVL dataset, which was released 2014. You might need to adjust the file formats to fit to the new dataset. For the NYU issues, you might take a look at issue #4

Hi， I have another question. When compare to other paper, I can only know their average joint error, but can not get their value on every kind joint, just a bar graph, how can I know their accurate value?

moberweger commented 5 years ago

Well, if you can get the coordinates of the joints of the other methods, you can plot all the different plots (see src/util/handpose_evaluation.py), otherwise you can only use the average joint errors for comparison.

zhanglong321 commented 5 years ago

Well, if you can get the coordinates of the joints of the other methods, you can plot all the different plots (see src/util/handpose_evaluation.py), otherwise you can only use the average joint errors for comparison.

Thanks for you answer, I got it!

zhanglong321 commented 5 years ago

Well, if you can get the coordinates of the joints of the other methods, you can plot all the different plots (see src/util/handpose_evaluation.py), otherwise you can only use the average joint errors for comparison.

Thanks for you answer, I got it!

Well, if you can get the coordinates of the joints of the other methods, you can plot all the different plots (see src/util/handpose_evaluation.py), otherwise you can only use the average joint errors for comparison.

Where can I get these coordinates data? I can't find it,