search

Nicholasli1995 / EvoSkeleton

Official project website for the CVPR 2020 paper (Oral Presentation) "Cascaded deep monocular 3D human pose estimation wth evolutionary training data"
https://arxiv.org/abs/2006.07778
MIT License
333 stars 44 forks source link

Is all of the mpi data used? #25

Closed gravitychen closed 3 years ago

gravitychen commented 3 years ago

happy spring festival,

I wonder about the details when you evaluate the model on the MPI-inf-3DHP dataset.

did you use all of the 8 subjects && 2 sequences && 14 cameras to inference the model?

Nicholasli1995 commented 3 years ago

MPI-inf-3DHP

Hi, we only use the testing set of MPI-INF-3DHP for evaluation. We did not use the training set for fine-tuning. We used the provided 2D keypoints as input to our pre-trained 2D-to-3D lifting model.

gravitychen commented 3 years ago

MPI-inf-3DHP

Hi, we only use the testing set of MPI-INF-3DHP for evaluation. We did not use the training set for fine-tuning. We used the provided 2D keypoints as input to our pre-trained 2D-to-3D lifting model.

Nice, so do you normalize MPI testing data using pre-saved stats.npy instead of calculating the mean and std of testing data

Nicholasli1995 commented 3 years ago

MPI-inf-3DHP

Hi, we only use the testing set of MPI-INF-3DHP for evaluation. We did not use the training set for fine-tuning. We used the provided 2D keypoints as input to our pre-trained 2D-to-3D lifting model.

Nice, so do you normalize MPI testing data using pre-saved stats.npy instead of calculating the mean and std of testing data

Yes. We use pre-computed stats.npy to normalize 2D joints of testing data.

gravitychen commented 3 years ago

OK I got it. thank you for your quick reply!