Closed ChenyuGao closed 5 years ago
Hi @ChenyuGao, Did you figure out how to rotate? I use the camera information from https://github.com/una-dinosauria/3d-pose-baseline , but the rotated result is wrong.
Use the function as follows which is provided by the author in another issue. Input is the pose parameters from neutrMosh/neutrSMPL_H3.6 dataset, then you can get the right pose as the original image (every 5 frames)
def rectify_pose(pose):
"""
Rectify "upside down" people in global coord
Args:
pose (72,): Pose.
Returns:
Rotated pose.
"""
pose = pose.copy()
R_mod = cv2.Rodrigues(np.array([np.pi, 0, 0]))[0]
R_root = cv2.Rodrigues(pose[:3])[0]
new_root = R_root.dot(R_mod)
pose[:3] = cv2.Rodrigues(new_root)[0].reshape(3)
return pose
@ChenyuGao Thanks.
do you have the neutrSMPL_H3.6 tar?can you send to me?my email address is yifengji1020@163.com,thanks
Hello,may I ask you for the neutrSMPL_H3.6 tar? I would be grateful if you could give it to me.My email address is chenjj954@163.com.
Unfortunately I cannot distribute that data anymore due to licensing reasons.
Thanks,
Angjoo
On Jun 13, 2019, at 8:26 PM, dangpugui notifications@github.com wrote:
Hello,may I ask you for the neutrSMPL_H3.6 tar? I would be grateful if you could give it to me.My email address is chenjj954@163.com.
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub, or mute the thread.
I am gold to receive your reply, does this mean that the training code will not work properly?
In fact, when I train without the H3.6M dataset, the code would always get stuck at this step.
---Original--- From: "akanazawa"notifications@github.com Date: 2019/6/14 12:12:40 To: "akanazawa/hmr"hmr@noreply.github.com; Cc: "dangpugui"153254904@qq.com;"Comment"comment@noreply.github.com; Subject: Re: [akanazawa/hmr] About the tf_records_human36m and neutrMosh/neutrSMPL_H3.6? (#75)
Unfortunately I cannot distribute that data anymore due to licensing reasons.
Thanks,
Angjoo
On Jun 13, 2019, at 8:26 PM, dangpugui notifications@github.com wrote:
Hello,may I ask you for the neutrSMPL_H3.6 tar? I would be grateful if you could give it to me.My email address is chenjj954@163.com.
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub, or mute the thread.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
Well the training code can still work, you just have to take out human3.6m from the command, something like:
CMD="python -m src.main --d_lr 1e-4 --e_lr 1e-5 --log_img_step 1000 --pretrained_model_path=${PRETRAINED} --data_dir ${DATA_DIR} --e_loss_weight 60. --batch_size=64 --use_3d_label True --e_3d_weight 60. --datasets lsp,lsp_ext,mpii,coco,mpi_inf_3dhp --epoch 75 --log_dir logs" But yes it will change the performance, should approach that of weakly-supervised setting (no explicitly 2D-to-3D pair) in the paper.
Best,
Angjoo
On Thu, Jun 13, 2019 at 11:54 PM dangpugui notifications@github.com wrote:
I am gold to receive your reply, does this mean that the training code will not work properly?
In fact, when I train without the H3.6M dataset, the code would always get stuck at this step.
---Original--- From: "akanazawa"notifications@github.com Date: 2019/6/14 12:12:40 To: "akanazawa/hmr"hmr@noreply.github.com; Cc: "dangpugui"153254904@qq.com;"Comment"comment@noreply.github.com; Subject: Re: [akanazawa/hmr] About the tf_records_human36m and neutrMosh/neutrSMPL_H3.6? (#75)
Unfortunately I cannot distribute that data anymore due to licensing reasons.
Thanks,
Angjoo
On Jun 13, 2019, at 8:26 PM, dangpugui notifications@github.com wrote:
Hello,may I ask you for the neutrSMPL_H3.6 tar? I would be grateful if you could give it to me.My email address is chenjj954@163.com.
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub, or mute the thread.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.
— You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub https://github.com/akanazawa/hmr/issues/75?email_source=notifications&email_token=AAHBXWBEHSQQEUUDLXU7DBTP2M6DVA5CNFSM4G4MCCVKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODXV4VTI#issuecomment-501992141, or mute the thread https://github.com/notifications/unsubscribe-auth/AAHBXWGJUUR2YQ72RZPA5I3P2M6DVANCNFSM4G4MCCVA .
Could you please send the neutrSMPL_H3.6 dataset and the tf_records_human36m dataset to me? They are necessary for reproducing the results. I would appreciate it if you can share them with me. Thanks a lot! My e-mail address is marvelzhy@gmail.com
do you have the neutrSMPL_H3.6 tar? can you send to me?my email address is [781256005@qq.com]
do you have the neutrSMPL_H3.6 tar? can you send to me?my email address is [474186834@qq.com]
Could you please send the neutrSMPL_H3.6 dataset and the tf_records_human36m dataset to me? They are necessary for reproducing the results. I would appreciate it if you can share them with me. Thanks a lot! My e-mail address is marvelzhy@gmail.com
hello, i also need the tf_records_human3.6M datasets for evaluation, if you have got it could you share with me, thank you very much! my email address is 1849535352@qq.com
Unfortunately the authors fo H36M do not want those files to be distributed, I recommend you folks to check out the newer version of HMR, which doesn't use any of this data and works much better: https://shubham-goel.github.io/4dhumans/
On Sat, Jun 3, 2023 at 12:30 AM Qusong1 @.***> wrote:
Could you please send the neutrSMPL_H3.6 dataset and the tf_records_human36m dataset to me? They are necessary for reproducing the results. I would appreciate it if you can share them with me. Thanks a lot! My e-mail address is @.***
hello, i also need the tf_records_human3.6M datasets for evaluation, if you have got it could you share with me, thank you very much! my email address is @.***
— Reply to this email directly, view it on GitHub https://github.com/akanazawa/hmr/issues/75#issuecomment-1574747409, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAHBXWAV2JFEUGO7AEM2VCDXJLR2LANCNFSM4G4MCCVA . You are receiving this because you modified the open/close state.Message ID: @.***>
-- Angjoo Kanazawa Assistant Professor Department of Electrical Engineering and Computer Sciences University of California, Berkeley https://people.eecs.berkeley.edu/~kanazawa/ She/Her
Dooray! 메일 발송 실패 안내
메일 발송 실패 안내
@.***) 님께 보낸 메일이 전송되지 못하였습니다.
실패 사유를 확인해보세요.
* 받는 사람 :
@.***)
* 발송 시간 :
2023-06-07T03:40:01
* 메일 제목 :
Re: [akanazawa/hmr] About the tf_records_human36m and neutrMosh/neutrSMPL_H3.6? (#75)
* 실패 사유 :
받는 사람의 메일함 용량이 가득 찼습니다. 받는 사람에게 메일함 정리를 요청하신 후 다시 발송해 보세요.
이 메일은 발신전용으로 회신되지 않습니다.
더 궁금하신 사항은
***@***.***
으로 문의해 주시기 바랍니다.
© Dooray!.
Could I please get the NeutrSMPL_H3.6 dataset for my research? It's really important for my work. My email is ildtechtop@gmail.com. Thanks a lot!!!
My question is: In the neutrMosh/neutrSMPL_H3.6 dataset and your processed tf_records_human36m, the SMPL pose's first 3 parameters are different. How to process it?
Below is what the author answered me: Quick answer is that those are two different mosh is in global coordinate space while the individual TF records for human 3.6m are in the camera coordinate frame. I think I rotated it using the provided camera information from human3.6m. You can also do it by solving the procrustes problem between the meshes to figure out the alignment.