search

intelligolabs / HARPER

HARPER is a HRI dataset for 3D Human Pose Estimation and Forecasting from the Robot’s Perspective.
https://intelligolabs.github.io/HARPER/
9 stars 0 forks source link

Baseline Code #1

Closed zhazhat closed 3 months ago

zhazhat commented 3 months ago

Hello, can you provide the baseline code for 3D pose estimation and motion prediction in the paper?

federicocunico commented 3 months ago

Dear @zhazhat thank you for reaching us out. The code for 2D human pose estimation is the HRNet, we used a pytorch implementation: https://github.com/leoxiaobin/deep-high-resolution-net.pytorch/tree/master/demo For the 3D pose estimation we lifted using the the HRNet + depth data (using the 'visibles_3d' key in the annotations). For the motion prediction, we used the official implementation of the methods (i.e. EQMotion).

We will release soon our validation code.

zhazhat commented 3 months ago

Ok, thank you for your quick answer! I would like to ask whether your method can achieve online prediction in real scenarios. I look forward to your work.

federicocunico commented 3 months ago

At the current state, no. This is a limitation we are currently investigating for future works. The 3D pose forecasting is technically able to do it (since they are approaches with "sliding window" behaviour). However, the reconstruction of the full pose requires the full sequence to re-create the skeleton (take a look at the paper to have a detailed explanation).

zhazhat commented 3 months ago

Ok, thanks for your answer