KDPE / traj_viz

Lightweight tooling to visualize robot rollout trajectories, images, depth, joint states and poses.

1. Prerequisites

Make sure you have:

• Git LFS (Large File Storage) (optional: to download the demo data).
• pixi

If you do not have pixi installed: https://pixi.sh/

2. Clone (with dataset via Git LFS)

git clone https://github.com/hsp-iit/KDPE.git
cd KDPE
git lfs install
git lfs pull

3. Environment Setup

Install dependencies (creates/updates the pixi environment defined in pyproject.toml):

pixi install

5. Visualizer API (Core Logic)

Visualizer (see src/traj_viz/visualizer.py) exposes:

Visualizer(
  urdf_path: str,
  robot_pose: Pose | None = None
)

Once constructed it creates an internal recording and logs a blueprint + URDF. You then call:

visualizer.log(
  joints: dict[str, np.ndarray] = {},         # joint_name -> scalar angle
  images: dict[str, np.ndarray] = {},         # name -> HxWxC (uint8 RGB)
  depths: dict[str, np.ndarray] = {},         # name -> HxW or HxWx1 float32 depth
  poses: dict[str, Pose] = {},                # name -> Pose(pos, ori, grip)
  cameras: dict[str, Camera] = {},            # intrinsic/extrinsic virtual cams
  trajectories: dict[str, np.ndarray] = {},   # name -> (T, >=6) (xyz + rotvec + optional scalar)
  timestamp: float = 0.0,
  static: bool = False,
)

6. Example Usage (inspect_rollout.py)

pixi run python src/examples/inspect_rollout.py