The code for the paper
Melica Omer Ali et. al. "EventPoseFormer: Event-based real-time 3D human pose estimation"
@inproceedings{ali2026,
title={EventPoseFormer: Event-based real-time 3D human pose estimation},
author={},
booktitle={venue},
pages={},
year={2026},
organization={IEEE}
}
Offline Execution
python evpf-offline.py --event_input_path "inputs/cam2_S1_Directions/ch0dvs" --output_dir "outputs" --skip 0-10 --save_images --save_demoOffline Evaluation
python 3d_evaluate.py --pred_path outputs/cam2_S1_Directions/predictions_3d.npz --subject S1 --activity Directions --frame 1000 --plotOnline Execution
- Install moveEnet and atis-gen-3 docker images from here: moveEnet and atis-gen-3
- Run the following command to start the moveEnet docker container:
docker run -it --privileged --network host -v /dev/bus/usb:/dev/bus/usb -v /tmp/.X11-unix:/tmp/.X11-unix -v yourpathtohpecore:/mnt/projects/hpe-core -e DISPLAY=unix$DISPLAY --name moveenet moveenet:latest- Inside the moveEnet docker container, run the following commands in separate terminals:
yarpservereros-framerNote: Run the following command to enter the moveEnet docker container from another terminal:
docker exec -it moveenet bash- In another terminal, run the following command to start the atis-gen-3 docker container:
lsusbUse the output of the above command to find the USB device ID of your ATIS camera. Then run:
docker run -it --network host --device /dev/bus/usb/[ID]:/dev/bus/usb/[ID] eventdrivenrobotics/atis-gen3- Inside the atis-gen-3 docker container, run the following commands:
yarp detect --writeatis-bridge-sdk --s 60 --filter 0.01- Run each of the following commands in separate terminals inside the moveEnet docker container:
python3 evpf-online.py --checkpoint_path /usr/local/hpe-core/example/movenet/models/e97_valacc0.81209.pthpython3 movenet_viewer.py --name /movenetVieweryarpview --name /viewerpython3 evpf-viewer1.py --name /evpfVieweryarp connect /atis3/AE:o /eroser/AE:i
yarp connect /eroser/img:o /movenet/img:i
yarp connect /eroser/img:o /movenetViewer/img:i
yarp connect /movenet/sklt:o /movenetViewer/sklt:i
yarp connect /movenetViewer/overlay:o /viewer
yarp connect /movenet/sklt3d:o /evpfViewer/sklt3d:i