Documentation

Data Download

Register an account here and then you will recieve a link to download the PALM dataset. After downloading, you will have this folder structure:

./PALM/XR20A/0000.zip
./PALM/XR20A/0001.zip
./PALM/XR20A/0002.zip
...
./PALM/XR20A/meshes/0000.zip
./PALM/XR20A/meshes/0001.zip
...
./PALM/XR20A/gt_checksums.json

Set environment variable, for example: export PALM_DIR=/media/zfan/ssd/palm_download/PALM/XR20A.

Create symbolic link:

cd code
ln -s $PALM_DIR PALMDB_highres

PALM Usage

To use PALM, you can use this simple script which gives an example to load the annotations and overlay the results on an image:

python visualize_PALM.py -c <cam_id> -id <subject_id> -frame <frame_num>

• cam_id is the id of the camera. All scans have 7 cameras with ids [MCU_01, MCU_02, MCU_03, MCU_04, MCU_05, MCU_06, MCU_07]
• subject_id is the id of the subject. There are a total of 262 subjects with ids [0000, 0001,...., 0262]
• frame_num Each scan has a varying number of frames in the range [0, 60]

Joint ordering

We use MANO joint ordering. See zc-alexfan/arctic for the joint names and ordering.