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
675 stars 92 forks source link

Running the video on the ReadME #148

Open anorakthegreat opened 9 months ago

anorakthegreat commented 9 months ago

On the ReadMe, there are a ton of videos showing real time and very accurate 3d pose estimation. Is the code for running that estimation public? I cant seem to find the code. Additionally, what model is the video using? Is that model public as well?

mks0601 commented 9 months ago

Hi, the videos are showing GT

anorakthegreat commented 9 months ago

Is that the model?

mks0601 commented 9 months ago

No they are GT

anorakthegreat commented 9 months ago

Sorry for the ignorant question, what is GT?

mks0601 commented 9 months ago

Groundtruth. It means annotation.

anorakthegreat commented 9 months ago

So, the videos on the README aren't the model/code, but just annotations? Again, sorry if this is a stupid question

mks0601 commented 9 months ago

yeap. it is to introduce our dataset, not to introduce our model. sorry for the confusion.

anorakthegreat commented 9 months ago

Ohhh so that video taken with all manual annotations? Is the model capable of accurately modeling videos?

mks0601 commented 9 months ago

no all machine annotation. there could be some errors for the models

anorakthegreat commented 9 months ago

Is there code for the machine annotation? And was that a no for the model being capable of labeling videos?

mks0601 commented 9 months ago

we do not release annotation code.

anorakthegreat commented 9 months ago

Is the pre-trained snapshot released full versions of the model, or do I need to train my own model with the provided dataset to get the best results?