lck1201 / simple-effective-3dpose-baseline Goto Github PK

How to predict my custom images, please?

General question the 44 mm error you give is that on the test set of on the training set?

When I run the code, it returns an error, how can I solve that?

Hi, many thanks for making this implement. I feel confused with the Human3.6m annotation you provided. Since the file size is almost twice larger than the original implement (https://github.com/una-dinosauria/3d-pose-baseline). So do they have the same quantity 3D annotations? Looks your annotation data has entire four camera perspectives for each pose, and so are they from the original Human3.6 dataset? Cause I don't have the access to download the Human3.6 data yet and have no idea what they look like. Many thanks!

When I test with your pre-training model, in line 52 of the main () function：logger. info ("Construct Dataset:", "config. DATASET. dbname [i]," "Dataset Path:", "config. DATASET. dataset_path [i]) will report errors: list index out of range. I can't find the reason for the mistake. Can you help me solve it? If you can, thank you very much.

From what I can tell the dataset you have uploaded is the 3d joint locations along with the corresponding camera intrinsics/extrinsics to perform perspective projection. Is this correct? If so is there anyway you could share the corresponding Euler Angle data?

hi:
First of all, congratulations on your good work. I am a graduate student inShanghai,At present, I am studying the direction of human posture detection, which requires the use of human 3.6m data set. Do you have any? Could you share it with me? Thank you for any reply @lck1201

It seems that your code use subject 1, 5, 6, 7 and 8 to compose training set, and use subject 9,11 to compose val set. But in Human 3.6M official document, the training set is composed of S1, S7, S8, S9 and S11 (the val set is composed of S5 and S6), so is there any particular reason to do this?

dear @lck1201,
Can you show the implement is more detail like the original?
The data is showed by chinese, can you update it in drive.google?