Ego Reconstruction Setup
SkillCloud & infraPrepare this repository’s egocentric reconstruction pipeline. Use when a user asks to install host packages, build Docker images, download model weights, verify Docker/GPU prerequisites, or diagnose missing setup for `reconstruction/modules/v2d_pipelines/run_ego_reconstruction.py`.
Available today. Use it from your connected AI after setup.
No other account needed.
Connect ahel once, and every AI you use reads what you have installed.
Then ask your AI: use the Ego Reconstruction Setup skill
What this skill tells your AI
The instructions your AI receives, as published by nvidia-isaac/video_to_data in .claude/skills/ego-reconstruction-setup/SKILL.md and read by ahel’s review.
Work from reconstruction/. Install the lightweight host wrappers, then build
only the pipeline images and download the matching weights:
bash scripts/install_ego_reconstruction_packages.sh
bash scripts/build_ego_reconstruction_packages.sh
bash scripts/download_ego_reconstruction_weights.sh --mode all
Select hamer_prompt, hamer_mesh, or dynhamr_prompt instead of all when
the run mode is known. hamer_mesh is for a caller-provided object mesh.
Before a long run, check the entrypoint and Docker/GPU availability:
python modules/v2d_pipelines/run_ego_reconstruction.py --help
docker version
nvidia-smi
Keep heavy ML dependencies in containers. For DynHaMR, verify the manual MANO
and BMC assets under data/weights/hand/ before running.
Signals
- GitHub stars
- 618
- Forks
- 60
- Last commit
- Sep 2026
Advanced
- Catalog kind
- skill
- Gateway key
ego-reconstruction-setup- Source
- github.com/nvidia-isaac/video_to_data