Question about Installation Instructions: Why `conda activate glamr` and How to Install System Dependencies Without `sudo`?

[Sjeary]

❗ Question about Installation Instructions: Why conda activate glamr and How to Install System Dependencies Without sudo?

Hi, thank you for releasing this great project.
While installing GLAMR following the README and install.sh, I encountered two issues that I hope to get clarification on.

---

1. Why does the installation guide suddenly use conda activate glamr?

In the README and the installation instructions, the environment name is never defined as glamr, but in the MOT installation step we see:

conda activate hybrik
pip install -e git+https://github.com/mkocabas/yolov3-pytorch.git#egg=yolov3
pip install -e git+https://github.com/haofanwang/multi-person-tracker.git#egg=multi-person-tracker
pip install numba filterpy
conda activate glamr

However:

• The README never instructs users to create an environment called glamr
• Only a hybrik environment appears in the dependency chain
• The meaning of "activate glamr" is unclear
• The glamr environment does not exist unless manually created by the user

❓ Is this a typo? Or should the user manually create a new environment for GLAMR?

If a separate environment is required, could you please provide:

• the correct environment creation command
• a clear indication of when to switch between environments

Right now, this part is confusing for new users.

---

2. Cannot run apt install — any recommended user-level alternatives?

The install.sh script requires system-level packages:

sudo apt install libgl1-mesa-glx xvfb
sudo apt update
sudo wget https://github.com/mmatl/travis_debs/raw/master/xenial/mesa_18.3.3-0.deb
sudo dpkg -i ./mesa_18.3.3-0.deb || true
sudo apt install -f

However, I do not have sudo privileges on my system (common on shared servers / institutional clusters), so it’s impossible to install these system packages.

Could you clarify:

• Which of these packages are strict requirements for GLAMR?
• Are there conda-based alternatives?
  (For example: mesalib, libglu, xorg-* packages from conda-forge.)
• Is it possible to fully run GLAMR without installing system-level Mesa?
• Does GLAMR actually require OSMesa for rendering, or only basic OpenGL libraries?

If you know a recommended workaround (e.g., conda substitutes, disabling GUI components, etc.), it would be extremely helpful for users who cannot run sudo apt.

---

Thanks again for the project — looking forward to your clarification!

---

[Sjeary]

Additional Question: What Dependencies Are Actually Required to Run GLAMR?

One more question regarding the installation process:

It seems that a large portion of the setup comes from the HybrIK environment (specific PyTorch versions, PyTorch3D, opendr, older SciPy, old OpenCV, etc.). Many of these dependencies are quite old or difficult to install (especially without root access).

❓ If the goal is just to run or reproduce the results of GLAMR (not training from scratch), which components are truly required?

More specifically:

• Do we really need the full HybrIK environment to run GLAMR’s demos?

• Which dependencies are required only for training, and which are required for inference / demo?

• Are modules like:

  • PyTorch3D
  • opendr
  • tb-nightly
  • system-level Mesa/OpenGL
  • yolov3 / multi-person-tracker

strict requirements for GLAMR, or optional depending on the task?

A minimal list of dependencies for simple inference (single-person or multi-person demo) would greatly help users reproduce your work without fighting numerous system and version conflicts.