---
title: EgoForce
emoji: 🐠
colorFrom: gray
colorTo: purple
sdk: gradio
sdk_version: 6.9.0
python_version: "3.10.13"
app_file: app.py
pinned: false
license: cc-by-nc-4.0
short_description: Forearm-Guided Camera-Space 3D Hand Pose
startup_duration_timeout: 1h
---

This Space runs the EgoForce video demo with the Hugging Face Gradio SDK and is prepared for ZeroGPU hardware.

At startup, `app.py`:

- Clones `https://github.com/dfki-av/EgoForce` into the Space runtime.
- Patches the upstream Gradio callback with `@spaces.GPU` so video inference runs inside ZeroGPU's dynamic CUDA allocation.
- Downloads the `_DATA` model assets from `chris10/EgoForce` on the Hugging Face Hub.
- Launches the upstream Gradio demo at `demo/run_app.py`.

Most Python dependencies are declared in `requirements.txt`, with build prerequisites in `pre-requirements.txt`. Packages that need custom pip flags or repo-local subdirectories are installed by `app.py` after the EgoForce source checkout. Debian packages needed for OpenCV, ffmpeg, headless rendering, and compiled extensions are declared in `packages.txt`.

Select `ZeroGPU` in the Space hardware settings. `ZEROGPU_DURATION_SECONDS` defaults to `600`, and `ZEROGPU_SIZE` defaults to `large`.