search

boycehbz / DMMR

The code for 3DV 2021 paper "Dynamic Multi-Person Mesh Recovery From Uncalibrated Multi-View Cameras"
MIT License
78 stars 6 forks source link

Experiments on MHHI datasets #14

Open chingswy opened 2 years ago

chingswy commented 2 years ago

Hello, thanks for sharing your code. I have two questions about the experiments on MHHI datasets:

  1. Sup. Mat. can't be found and I want to know the details of this dataset.

    The details of the datasets that are used for training and testing can be found in the Sup. Mat.

  2. How to select the paired 3D vertices of the tracked markers.

    The numbers are the mean distance with standard deviation between the tracked 38 markers and its paired 3D vertices in mm.

boycehbz commented 2 years ago

Hi, thanks for your interest.

  1. The Sup. Mat. can be found at here.

  2. We use the evaluation code provided by Li et al. to calculate the vertex error, which can be found at here.

chingswy commented 2 years ago

Thanks! It seems that only doubleB sequence is used in evaluation. Can we use the provided reconstructed mesh of other sequences for evaluation? Are they accurate enough?

boycehbz commented 2 years ago

Only the doubleB in MHHI dataset is captured with mocap system, and I found the provided ground-truth 3D joints in other sequences are inaccurate. So, we follow previous works and use doubleB only for evaluation.