search

facebookresearch / InterHand2.6M

Official PyTorch implementation of "InterHand2.6M: A Dataset and Baseline for 3D Interacting Hand Pose Estimation from a Single RGB Image", ECCV 2020
Other
676 stars 92 forks source link

Have you updated all your images, lables and MANO parameters in September? #84

Open MengHao666 opened 2 years ago

MengHao666 commented 2 years ago

I downloaded images, lables and MANO parameters half year ago, but now I could see the dates of file in your google drive are updated in Septmeber. I wonder if I should re-download all these things again? Especially, MANO parameters are very crucial to me. Did you update it to be more accurate?

MengHao666 commented 2 years ago

And it seems that the name rule of MANO parameters, that is, you might integrate all MANO parameters that is of same mode but different annot_subset into one file. Did you modify the content inside?

mks0601 commented 2 years ago

Images and labels are updated, but I just added some new fields. GT 3D joint coordinates haven't updated. MANO parameters might be updated, but can't remember clearly :( Could you check by yourself?

MengHao666 commented 2 years ago

Why don’t you rename a new version for this update? For fair comparison,could I use the not updated data? But,what do you mean by updating images and labels, and adding some fields?The old dataset has some errors in annotation.For example ,a interacting hand might not contain both annotations. Do you mean you annotate again for such samples?And I use 5fps version. It would be better to check it yourself, as you can give an official conclusion.---- Replied Message ----FromGyeongsik @.>Date10/06/2021 00:26 @.> CcHao @.**@.>SubjectRe: [facebookresearch/InterHand2.6M] Have you updated all your images, lables and MANO parameters in September? (#84) Images and labels are updated, but I just added some new fields. GT 3D joint coordinates haven't updated. MANO parameters might be updated, but can't remember clearly :( Could you check by yourself?

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.

[ { @.": "http://schema.org", @.": "EmailMessage", "potentialAction": { @.": "ViewAction", "target": "https://github.com/facebookresearch/InterHand2.6M/issues/84#issuecomment-934565046", "url": "https://github.com/facebookresearch/InterHand2.6M/issues/84#issuecomment-934565046", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { @.": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

mks0601 commented 2 years ago

I confirmed that training models with annotations of previous versions and current versions produce almost the same results when testing it on both previous/current annotations. The newly added field is 'seq': str (sequence name) in joint_3d.json

MengHao666 commented 2 years ago

Thanks so much for explanation. It woul not influence results considering only new field is added and there are no additional images. Is my understanding true?And how about MANO parameters?Did you update them to better ones?---- Replied Message ----FromGyeongsik @.>Date10/06/2021 01:24 @.> CcHao @.**@.>SubjectRe: [facebookresearch/InterHand2.6M] Have you updated all your images, lables and MANO parameters in September? (#84) I confirmed that training models with annotations of previous versions and current versions produce almost the same results when testing it on both previous/current annotations. The newly added field is 'seq': str (sequence name) in joint_3d.json

—You are receiving this because you authored the thread.Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.

[ { @.": "http://schema.org", @.": "EmailMessage", "potentialAction": { @.": "ViewAction", "target": "https://github.com/facebookresearch/InterHand2.6M/issues/84#issuecomment-934611257", "url": "https://github.com/facebookresearch/InterHand2.6M/issues/84#issuecomment-934611257", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { @.": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

mks0601 commented 2 years ago

There is no new image. Sorry for the MANO parameters, but I can't remember clearly. I guess there could be some minor updates, but not major updates though.

MengHao666 commented 2 years ago

There is no new image. Sorry for the MANO parameters, but I can't remember clearly. I guess there could be some minor updates, but not major updates though.

@mks0601 Thanks fo reply. But, did the minor updates of mano parameters get lower fitting error than the original one? Could you provide quantitive results abput this? I guess if the fitting result is worse, you don't need to update it. And I have checked the new MANO parameters are different from original one, but I don't know which one to use.